# Papers by Keyword ### Agent / LLM Agent - RecSys Factory: Bounding LLM Agent Autonomy to Decision Points in the Industrial Recommender Lifecycle (RecSys Factory) — Tencent, arXiv 2608.11241 - PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems (PsyBer-Agent) — XJTLU / U Liverpool / Alibaba, arXiv 2610.13989 - HEM-Med: A Hierarchical Agentic Memory Framework for Safe Multi-Center Medication Recommendation (HEM-Med) — Qilu University of Technology / UTS, arXiv 2610.20191 - Advancing Model Research in AgentX: Long-Horizon Autonomy for Industrial Recommender Systems (AgentX-Model) — Kuaishou, arXiv 2609.30001 - Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops — Spotify, arXiv 2609.30297 - Effective Reinforcement Learning for Agentic Search by Recycling Zero-Variance Queries During Training - ANCHOR: Agentic Noise Creation Framework for Human Simulation and Denoising Recommendation (ANCHOR) - AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems (AgentX) - MM-VeriRec: Failure-Guided Fusion for Verifiable Agentic Multimodal Recommendation — Independent Researcher, arXiv 2609.31718 - NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems (NOVA) - On Evaluating and Improving Conversational Agents in Production — Zalando, arXiv 2609.32092 - Prompt Optimization for User Simulation in Conversational Recommender Systems (UserSimulator) - tau-Rec: A Verifiable Benchmark for Agentic Recommender Systems - Skill Is Not Document: A Query-Conditional Benchmark and Two-Stage Retriever for LLM Agent Skill Routing (R3) - VirtualMLE: A Virtual ML Engineer that Optimizes Sequential Recommenders (VirtualMLE) - Self-Distilled Reinforcement Learning for Co-Evolving Agentic Recommender Systems (CoARS) - Let the Agent Steer: Closed-Loop Ranking Optimization via Influence Exchange (Sortify) - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - EvoRec: Self Evolving Agentic Recommender Systems (EvoRec) - ReasonRec: A Reasoning-Augmented Multimodal Agent for Unified Recommendation (ReasonRec) - RouteRec: Strict Evaluation of Recommender-Agent Selection and Aggregation (RouteRec) - RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems (RecHarness) - SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents (SafeGEO) - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - CoPersona: Collaborative Persona Graphs for Robust LLM Personalization (CoPersona) - ExPerT: Personalizing LLM Responses to Users' Domain Expertise via Query-Wise Semantic and Keystroke Behavioral Cues (ExPerT) - AgentSelect: Benchmark for Narrative Query-to-Agent Recommendation (AgentSelect) - Autonomous Information Seeking: A Roadmap for Agentic Recommender Systems - Seeing and Reflecting: Multimodal Memory-Enhanced Agent Collaboration for Recommendation (MMEACR) - An LLM-powered Agentic Recommendation System for Connected TV Content Discovery - HyCoRec: Hypergraph-Enhanced Multi-Preference Learning for Alleviating Matthew Effect in Conversational Recommendation (HyCoRec) - Mitigating Matthew Effect: Multi-Hypergraph Boosted Multi-Interest Self-Supervised Learning for Conversational Recommendation (HiCore) - ItemRAG: Item-Based Retrieval-Augmented Generation for LLM-Based Recommendation (ItemRAG) - SimGR: Escaping the Pitfalls of Generative Decoding in LLM-based Recommendation (SimGR) - STEPS: A Self-Triggered Agentic Push Recommendation System (STEPS) - TwiSTAR: Think Fast, Think Slow, Then Act — Generative Recommendation with Adaptive Reasoning (TwiSTAR) - Personalized Communication Skills for Agentic Recommender Systems (AgentCom) - MACS: A Hybrid Multi-Agent Framework for Reliable Conversational E-Commerce Recommendation (MACS) - When Memory Takes Gradients: Collaborative Vector Memory for Agentic Recommender Systems (CoVeMem) - Astar: Learning to Propose Evolution Directions for Self-Evolving Industrial AI Systems (Astar) - DREAMS: Dual-node Monte Carlo Tree Search for Conversational Recommendation (DREAMS) - RecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender Systems (RecEvolve) - Agents as Knowledge Integrator and Utilizer in Multimodal Recommendation (AgentMMRec) - AtomRec: Evolving Atomic Memory for Agentic Recommendation (AtomRec) - AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems (AutoLR) - Agentic ML Exploration (A-MLE) for Ads Ranking (A-MLE) - An Efficient and Effective Agentic Group Shilling Attack on Recommender Systems (AGAS) - Auto-RecSys: Harnessing Autonomous Research Agents for Industry-Scale Recommender System (Auto-RecSys) - Agentic Share-of-Search: A Multi-Agent AI System for Competitive Decision-Making in LLM-Mediated E-Commerce (ASoS) - Who Are We Recommending To? Recommender Systems in the Agentic Web (Spotify) — RecSys 2026 - Position: Recommender Systems Should Move Beyond Platform-Centric Ranking toward Personal Agent-Mediated Recommendation (PAMR) - MemRetriever: Learning to Search, Reflect, and Retrieve from Long-Term Memory (MemRetriever) - AURA: Agentic Diagnosis and Refinement for Production Recommender Systems at Scale (AURA) - LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory, with ICE v2 as an Audited Local-First Architecture (LSREP) - SELF-INDEX: Self-Evolving Search Index (SELF-INDEX) — Yonsei University (with Samsung Research, UC Irvine, Korea University), arXiv 2609.19656 - CORAL: An LLM-Native Harness for Production Recommender Systems (CORAL) — Meta AI, arXiv 2609.02730 - Enhancing Group Recommendation with Memory-Augmented Reasoning in LLM Agent (AGR) - Weather- and Location-Aware Agentic Dining Recommendation (Agentic Dining) — Independent; arXiv 2608.07593 - RecToolBench: Benchmarking Recommendation-Specific Tool Orchestration under Fuzzy User Intent (RecToolBench) — University of Virginia / Jilin University / Squirrel AI / PolyU, arXiv 2609.30717 - AgentRecommender: LLM Agents Enable Customizable Recommender Systems on the User Side (AgentRecommender) — National Institute of Informatics, arXiv 2609.31166 - ReMem: Multi-memory GRPO for Long-Context Recommendation Agents (ReMem) - Hong Kong Polytechnic University / NTU, arXiv 2609.37311 - RankEvolve: A Reliable Multi-Agent Auto-Research Harness for Evolving Ranking Models (RankEvolve) — Meta, arXiv 2609.39551 - The Like Trap: Multi-Stage Poisoning against Agents in Similarity-based Recommendation Systems (Like Trap) — Michigan State University, arXiv 2609.27155 - AgentWebRec: Compact Evidence Fusion over the Agent Web for Personalized Recommendation (AgentWebRec) — Beihang University, arXiv 2610.01705 - Personalization Matters: Long-Horizon Conversation Agent with User-Centric Information in Online Shopping Interactions (Personalization Matters) — ByteDance / University of Melbourne, arXiv 2610.11375 - Personal-Agent Mediated Recommendation with Cross-Platform User History (MediateRec) — UC San Diego / Meta (PAMO policy-optimization mediation), arXiv 2610.07588 - Delegation Asymmetry in Agentic Recommender Systems: Measuring Two-Sided Receptivity in Online Dating — University of Notre Dame / Fleamily, arXiv 2608.18058 - UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents (UserArena) — Leiden University / Renmin University of China / Alibaba Group / Institute of Computing Technology, CAS, arXiv 2610.30932 [opensource] ### Beam Search / Constrained Decoding - ReGen: Controllable Generative Recommendation via Guided Token Refinement (ReGen) — XJTLU / U Liverpool / PolyU, arXiv 2610.28642 [opensource] - Generative Recommenders Forget the Long Tail: Restoring Memorization in Retrieval and Ranking (CAPE) — USTC & Ant Group, arXiv 2610.29298 - FedCGR: Federated Cross-Domain Generative Recommendation (FedCGR) — CIKM 2026 - GenRec: An LLM-Backed Recommendation Ranker at Netflix (GenRec) - GRACE: Generative Recommender Acceleration Engine for Real-Time Ads Retrieval (GRACE) - HRPO: Hierarchical Residual Policy Optimization for Generative Recommendations (HRPO) - SITA: Semantic Interest Tokens for Target-Aware Compression in Long-Sequence Recommendation (SITA) - SmartGR: Hierarchy and Beam-Aware Knowledge Distillation for Generative Recommendation (SmartGR) - UniR²: Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence (UniR²) - Gryphon-v2: One Model in Place of a Cascade — Generate-and-Rank Recommender with Rollout Distillation (Gryphon-v2) - UniGD: A Unified Generative-Discriminative Framework for Industrial Retrieval (UniGD) - PinRec: Unified Generative Retrieval for Pinterest Recommender Systems (PinRec) - SA²CRQ: Towards Efficient and Generalizable Retrieval — Adaptive Semantic Quantization and Residual Knowledge Transfer (SA²CRQ) - OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation (OneLive) - SID Understanding: Understanding Semantic IDs — From Item Representation to Item Selection in Generative Recommendation (SID Understanding) - HD-Rec: Hierarchical Quantization with Domain-Adaptive Sparse Routing for Generative Cross-Domain Recommendation (HD-Rec) - TM20K: Teacher Retains Full Tokens, Student Merges Efficiently — E-Commerce Sequence Modeling in Ad Recommendation (TM20K) - Preserving Item Semantics for Free: Rethinking Token Initialization in LLM-Based Generative Recommendation (Snap/UMich) - PushDualGen: Enabling LLMs to Generate Semantic IDs with Interpretable Copy for Industrial Push Recommendation (Kuaishou) - MetaStrategy: Generative Ranking with Executable LLM Strategies (Alibaba Taobao) - TSPORec: Token Selection via Preference Optimization for LLM-Based Sequential Recommendation (ZJU/ByteDance) - IntHQ: Task-Interactive Hierarchical Query on Dual-Stream Representations for GenRec (Amap/Alibaba) - InforID: Adaptive Semantic Capacity Allocation for Parallel Generative Recommendation (UCAS/CASIA) - HCGRec: Hint-Conditioned Generative Recommendation with Semantic IDs (HCGRec) - Token-Level Credit Assignment Optimization for Generative Document Retrieval - TGR: Tencent Generative Recommendation — Unified Generation and Reasoning (TGR) - hLLM: Single Pass Decoding for Generative Reranking (hLLM) - WIDE: Wildcard Inference with Dynamic Expansion for Cross-Modal Generative Retrieval (WIDE) - OneLA: Scaling Linear-Attention Decoding to Large Beams in Generative Recommendation (OneLA) - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 ### CTR Prediction - FineSID: Progressive Learning from Content to Relevance to Personalization for CTR Prediction (FineSID) — Alibaba, arXiv 2610.29138 - DeRes: Decoupling Residual Stability and Adaptivity for Scalable CTR Prediction (DeRes) - Dual-Stream MLP is All You Need for CTR Prediction (DS-MLP) - Fine-Tuned LLM as a Complementary Predictor Improving Ads System - GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling (GateDiffInt) - Generative Long-term User Interest Modeling for Click-Through Rate Prediction (GenLI) - Joint Optimization of Relevance and Engagement in Multi-Task Ranking for E-Commerce with Efficient LLM Supervision - LLM-HYPER: Generative CTR Modeling for Cold-Start Ad Personalization via LLM-Based Hypernetworks (LLM-HYPER) - LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction (LoopCTR) - SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders (SORT) - TMallGS: Scaling Unified Feature and Sequence Modeling for Generative E-commerce Search (TMallGS) - Cheaper is Better: A Discount-Aware Network for Conversion Rate Prediction in E-commerce Recommendation System (DANet) - STAR: Structured Tokenization and Target-Aware Interest Representation for PCVR Prediction (STAR) - UniDot: A Unified Network for Sequence Modeling and Feature Interaction in Large-scale Recommendation (UniDot) - Cascading Relevance-driven Recommendation Network for CTR Prediction in Trigger-Introduced Recommendation (CRRN) - Native Multimodal Representation Learning for Click-Through Rate Prediction in E-Commerce Scenarios - DCEO: Direct Causal Effect Optimization for Long-Term User Value Modeling in E-commerce Search (DCEO) - Topology-Masked Unified Backbone for Joint Feature Interaction and Multi-Domain Sequence Modeling (MaskRec) - Stageboost: Recommending Signals Based on Counterfactual Estimation - HubMixer: Progressive Latent Hub Mixing for Parameter-Efficient Feature Interaction in Recommendation (HubMixer) - From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs (ReST) - From Feature Interaction to Feature Transport - A Unified Block for Scalable Recommendation Models (CRAFT) - UniCon: A Unified Context-Centric Modeling Paradigm for CTR Prediction (UniCon) - Personalized Task Dependency Graphs for Mitigating Signal Erosion in Multi-Task Recommendation (PTDG) - SequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching (SequenceO1) - Dense Feature Representation over Sequence Modeling: A Solution to the KDD Cup 2026 UniRec Challenge (KDD Cup 2026 UniRec) — Z Lab (Chengdu), KDD Cup 2026 Tencent UniRec Challenge Workshop - RoleMix: Unifying Sequential and Non-Sequential Features via Semantic Tokenization for Post-Click Conversion Rate Prediction (RoleMix) — UESTC / Macau Polytechnic University / University of Liverpool, KDD Cup 2026 Tencent UniRec Challenge, arXiv 2607.22700 ### Cold-Start - Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling - Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation (Shallow-RHS) - Cold-Start Personalization via Training-Free Priors from Structured World Models (PEP) - DiffCold: A Diffusion-based Generative Model for Cold-Start Item Recommendation (DiffCold) - Mind the Gap: Bridging Behavioral Silos with LLMs in Multi-Vertical Recommendations - POI Recommendation with LLM-Augmented Multi-Graph Learning and Contrastive Alignment (LLM-MGCL) - Representation Curriculum: Stagewise Training for Robust Ranking and Allocation - DREAM: Dynamic Refinement of Early Assignment Mappings (DREAM) - Denoising Implicit Feedback for Cold-start Recommendation (DIF) - VCG: A Multimodal Retrieval Framework for E-Commerce Video Feeds under Extreme Cold-Start Conditions (VCG) - GenRecEdit: Adapting Model Editing for Generative Recommendation with Cold-Start Items (GenRecEdit) - Diagnosing and Mitigating Retrieval Bottlenecks in LLM-Based Cold-Start Recommendation (LHF) - Serving the Long Tail: Training-Free LLM Candidate Generation for Vacation Rental Marketplaces - Learning Sparse Representations of Multimodal Content for Enhanced Cold Item Recommendation (QMUL, RecSys 2026) - Can Generative Recommendation Reach Cold Items? A Temporal Perspective on Semantic-ID Generation (ColdSID) - ItemRAG: Item-Based Retrieval-Augmented Generation for LLM-Based Recommendation (ItemRAG) - PinEqualizer: Full Funnel Content Exploration and Debiasing System at Pinterest (PinEqualizer) - VaLiDRec: Variable-Length LLM-Aligned Semantic IDs for Generative Recommendation (VaLiDRec) - MOTIF: Motivation-guided Topology Inference for Cold-start Multimodal Recommendation (MOTIF) - Personalized and Multi-View Representation for Federated Cold-Start Recommendation (PMFRec) - SwapRec: Warming Up Cold Items Through Training-Time Swaps (SwapRec) - Single-Token Expected-Value Scoring for Cold-Start Candidate Ranking - A Redundancy Reduction Approach for Controllable Sequential Recommendations (BT-SR) — Yandex / AIRI / HSE; Barlow-Twins decorrelation for head/tail exposure control - DualGCN: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation (DualGCN) — HKUST (Guangzhou), arXiv 2610.19578 ### Concept Unlearning - Neighborhood-Aware Unlearning for Generative Recommendation (NAU) — BIFOLD & TU Berlin / University of Amsterdam, arXiv 2610.27232 - TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation (TRACER) - SURF: Subtractive Updates for Recommender Forgetting (SURF) ### Contrastive Learning - GCIB: Graph Contrastive Information Bottleneck for Multi-Behavior Recommendation - Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation (ReaEmb) - POI Recommendation with LLM-Augmented Multi-Graph Learning and Contrastive Alignment (LLM-MGCL) - Quality-Aware Collaborative Multi-Positive Contrastive Learning for Sequential Recommendation (QCMP-CL) - UFRec: Uncertainty-Guided Future Learning for Sequential Recommendation ### Conversational - Dialogue Generation for Conversational Recommendation Using User Search Interactions and Knowledge Graphs (SearCl) — KAIST & LG Electronics, arXiv 2610.25182 - BanglaShop-CRS: A User-Centric Bangla Dataset for Conversational Recommendation (BanglaShop-CRS) — University of Vermont - Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops — Spotify, arXiv 2609.30297 - On Evaluating and Improving Conversational Agents in Production — Zalando, arXiv 2609.32092 - Overview and Analysis of the RecSys Challenge 2026: Conversational Music Recommendation — Sony Group Corporation, arXiv 2609.33045 - A Standardized Re-evaluation of Conversational Recommender Systems on the ReDial Dataset - GCRS: Generative Conversational Recommender System - MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music - Prompt Optimization for User Simulation in Conversational Recommender Systems (UserSimulator) - Towards Fast Domain Adaptation and Fine-Grained User Simulation for Evaluating Conversational Recommender Systems (AdaptSim) - TRACE: A Conversational Framework for Sustainable Tourism Recommendation with Agentic Counterfactual Explanations - Dialogue to Discovery: Attribute-Aware Preference Elicitation for Conversational Product Search Assistants (D2D) - Evaluating Scene-based In-Situ Item Labeling for Immersive Conversational Recommendation (ICRS) - When and How to Ask: Dynamic Preference Elicitation Strategies for Conversational Recommendation (COPE) - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation (Tencent/ACL 2026) - Ask to Be Sure: Informative Interactions for Confident Multi-Turn LLM Recommendation (Ask to Be Sure) - MACS: A Hybrid Multi-Agent Framework for Reliable Conversational E-Commerce Recommendation (MACS) - Conversational Recommendation over Live E-Commerce Catalogues with Self-Refreshing Retrieval - DREAMS: Dual-node Monte Carlo Tree Search for Conversational Recommendation (DREAMS) - Purchase Advice and Observable Buyer Responses in Real AI Conversations - FacetCRS: Multi-Faceted Preference Learning for Pricking Filter Bubbles in Conversational Recommender System (FacetCRS) — Sun Yat-sen University, AAAI 2024 - Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment (Re2A) — HK PolyU / CUHK / Sichuan University, EMNLP 2026 - Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation (CRS-Performance) - When the Label Ignores the Request: Auditing Policy-Selected Targets in Synthetic Conversational Music Recommendation (When-the-Label) — Uber AI, arXiv 2609.39696 [opensource] - Adapting Generative Recommenders for Multi-Turn Interaction (INTEGER) — National Taiwan University / Johns Hopkins University, arXiv 2610.08136 - Personalization Matters: Long-Horizon Conversation Agent with User-Centric Information in Online Shopping Interactions (Personalization Matters) — ByteDance / University of Melbourne, arXiv 2610.11375 - Shape Your Feed: An LLM-based Agentic System for Conversational Recommendation (SYF) — Meta Platforms, arXiv 2608.06632 - JobCCC: A Constraint-Aware Conversational Job Recommendation Benchmark and Method (JobCCC) - East West University / UC San Diego, arXiv 2610.05787 [opensource] - SafeCRS: Personalized Safety Alignment for LLM-Based Conversational Recommender Systems (SafeCRS) — UIC / UIUC / Amazon, KDD 2026 — arXiv 2603.03536 [opensource] - RAR: Retrieval Augmented Conversational Recommendation with Reinforcement Learning (RAR) — UIUC / Google DeepMind / UC San Diego — arXiv 2604.04457 [opensource] ### Cross-Domain - ATLAS: Learning to Recommend Across Unseen Domains (ATLAS) — Sony Research India, arXiv 2608.03899 - Cross-Country Code-Mixing for Generative Recommendation (CMRec) - Breaking the Information Silo: Semantic Personas for Cross-Domain Recommendation (SPHERE) - Bridging Short Videos and Live Streams: Reasoning-Guided Multimodal LLMs for Cross-Domain Representation Learning (RGCD-Rep) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - Mind the Gap: Bridging Behavioral Silos with LLMs in Multi-Vertical Recommendations - Atomic Intent Reasoning: Bringing LLM Semantics to Industrial Cross-Domain Recommendations (AIR) - Quantizing Intent: Cross-Domain Semantic IDs from Organic Activity for Industrial Ranking - One Sequential Recommendation Model Pretrained from Synthetic Priors Predicts Multiple Datasets (SRPFN) - Synthetic Data from Cross-Domain Events for Large-Scale Recommendation Systems (SCALR) - Causal-Invariant Cross-Domain Out-of-Distribution Recommendation (CICDOR) - HD-Rec: Hierarchical Quantization with Domain-Adaptive Sparse Routing for Generative Cross-Domain Recommendation (HD-Rec) - FedCGR: Federated Cross-Domain Generative Recommendation (FedCGR) — CIKM 2026 - A Dual-Expert Strategy Integrating LLMs to Mitigate Negative Transfer in Cross-Domain Sequential Recommendation (DuELRec) - Empowering Cross-Domain Sequential Recommendation with Hybrid Tokenization and Serial-Parallel Decoding (GenCDSR) - Reading the Mood: Emotion-Guided Book-to-Music Recommendation via CGANs and LLMs (SAGA-CDR) — National Technical University of Athens, SENTIRE 2026 (ICDM Workshops), arXiv 2610.06703 - When Semantics Fall Short: Persona-Guided Cross-Domain Sequential Recommendation with Large Language Models (PerCDR) — Beijing Jiaotong University, arXiv 2610.25221 - BGAFlow: Bezier-Guided Adaptive Flow for Cross-domain Sequential Recommendation (BGAFlow) — Anhui University, arXiv 2610.30825 ### Data Generation / Synthetic Data - From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations - One Sequential Recommendation Model Pretrained from Synthetic Priors Predicts Multiple Datasets (SRPFN) - Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation (Scaling Laws) - RSIR: Can Recommender Systems Teach Themselves? A Recursive Self-Improving Framework with Fidelity Control (RSIR) - SAGE (Tabular Data Generation) - World Model-Guided Reinforcement Learning via Counterfactual User Engagement Simulation (WMG-RL) ### Denoising - Adapting Knowledge Graphs for Behavior Denoising in Sequential Recommendation (AdaptedKG) - ANCHOR: Agentic Noise Creation Framework for Human Simulation and Denoising Recommendation (ANCHOR) - DC4SR (Disagreement as Signals) - Denoising Implicit Feedback for Cold-start Recommendation (DIF) - When Recommendation Denoising Meets Popularity Bias: Understanding and Mitigating Their Interaction (PAD) - GroundedGEO: Auditing the Evidence Gap in Generative Search Rankings (GroundedGEO) — Shenzhen University, arXiv 2609.25189 ### Diagnostics / Analysis - LIME-Rec: Auditing Semantic Gains in Sequential Recommendation (LIME-Rec) - SIDScope: A Diagnostic Resource for Semantic-ID Interfaces in Generative Recommendation (SIDScope) - Residual Dominance as a Structural Account of Last-Item Reliance in Causal Self-Attention Recommenders (Residual Dominance) - The Disconnect Between Better Descriptive Reasoning Trace Quality and Recommendation Effectiveness - Auditing Return Conditioning as a Control Knob: An Offline Diagnostic for Decision Transformer Recommendation - When Does Supervised Fine-Tuning Reduce Instruction Sensitivity? - Purchase Advice and Observable Buyer Responses in Real AI Conversations - On the Regularization Landscape for the Linear Recommendation Models - AURA: Agentic Diagnosis and Refinement for Production Recommender Systems at Scale (AURA) - LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory, with ICE v2 as an Audited Local-First Architecture (LSREP) - How Calibration Content Shapes Attention-Based Reranking (Interpolated Null Calibration) - Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding Learning (CoFree) — Alibaba (Taobao) / Wuhan University / Shanghai Jiao Tong University, arXiv 2609.20563 - SELF-INDEX: Self-Evolving Search Index (SELF-INDEX) — Yonsei University, arXiv 2609.19656 - Epistemic Warrant for LLM Recommendations: Characterizing the Basis for Reliance When Ground Truth Is Unavailable — Purdue University / University of Pennsylvania, arXiv 2609.04127 - When More Data Is Not Enough: The Context-Sufficiency Frontier in Generative AI Personalization (Context-Sufficiency) — ABYAT, arXiv 2610.00654 ### Diffusion - Anatomy of a Decision: Uncertainty-aware Hierarchical Intent Learning via Flow Matching for Multimodal Recommendation (UHIFlow) — Northeastern University, China, arXiv 2609.29609 - X-Rec Technical Report (X-Rec) — TikTok, arXiv 2609.29180 - GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling (GateDiffInt) - MIRAGE: Escaping the Euclidean Void (Manifold-Informed Flow Matching for Sequential Recommendation) - Adaptive Autoguidance for Item-Side Fairness in Diffusion Recommender Systems (A2G-DiffRec) - DiffCold: A Diffusion-based Generative Model for Cold-Start Item Recommendation (DiffCold) - Diffusion-GR2: Diffusion Generative Reasoning Re-ranker (Diffusion-GR2) - Dual-Diffusional Generative Fashion Recommendation (DualFashion) - FAVE: Flow-based Average Velocity Establishment for Sequential Recommendation - From Noise to Order: Learning to Rank via Denoising Diffusion (DiffusionRank) - Generative Retrieval via Diffusion Transformer with Metric-Ordered Sequence Training and Hybrid-Policy Preference Optimization (MO-DiT+HPPO) - Interests Burn-down Diffusion Process for Personalized Collaborative Filtering (StageCF) - LLaDA-Rec: Discrete Diffusion for Parallel Semantic ID Generation in Generative Recommendation - MDGR: A Masked Diffusion Generative Recommendation Framework (MDGR) - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - Time-Aware Diffusion based on Preference Disentanglement for Generative Recommendation (TDPM) - Diffusion Language Model for Recommendation (DLMRec) - TimeRoute: Time-Aware Modality Routing and Diffusion for Multi-Modal Recommendation (TimeRoute) - Adaptive Item-based Collaborative Structures via Noise Rescheduling in Diffusion for Generative Recommendation (ANR-DiffRec) - Preference Flow Matching with Spectral Factorization for Micro-video Recommendation (PrismRec) - EPIC: Explicit Posterior Item Conditioning for Semantic ID Diffusion Recommendation (EPIC) - MGDiff: Multi-Interest Sequence Recommendation with Masking GNN-Guided Diffusion (MGDiff) - PAPA: Online Personalized Active Preference Alignment (PAPA) — Washington University in St. Louis, arXiv 2607.00486 - Eval4DiRec: A Unified Evaluation Framework for Diffusion-based Recommender Systems (Eval4DiRec) - University of Technology Sydney, ACM TKDD, arXiv 2609.34404 - FairDiff: Mitigating the Self-Reinforcing Matthew Effect in Diffusion Recommender Models (FairDiff) — Tsinghua University / Huawei Noah's Ark Lab / USTC, arXiv 2609.36671 - Disentangling Paradigm, Identifier, and Decoding in Generative Retrieval (Disentangling Paradigm/Identifier/Decoding) — Artefact Research Center / Université d'Angers (LERIA), arXiv 2610.08716 - JBM-Diff: Joint Behavior-guided and Modality-coherence Conditional Graph Diffusion Denoising for Multi-Modal Recommendation (JBM-Diff) — Huazhong University of Science and Technology, arXiv 2604.03654 - FatsMB: From Agnostic to Specific: Latent Preference Diffusion for Multi-Behavior Sequential Recommendation (FatsMB) — CAS / Kuaishou, arXiv 2602.23132 - DiffSBR: Diffusion-based Latent Neighbor Generation for Session-based Recommendation (DiffSBR) — UESTC, arXiv 2601.03903 - DiffuReason: Bridging Latent Reasoning and Generative Refinement for Sequential Recommendation (DiffuReason) — Tencent (Think-then-Diffuse + GRPO), arXiv 2602.09744 - CARD: Counterfactual Attention Regulation Diffusion for Diffusion-based Sequential Recommendation (CARD) — Sun Yat-sen University / Peng Cheng Laboratory, ICASSP 2026 — arXiv 2601.15673 [opensource] - DiffGRM: Diffusion-based Generative Recommendation Model (DiffGRM) — Kuaishou, WWW 2026 (masked discrete diffusion for SID generation) — arXiv 2510.21805 [opensource] - OT-FlowRec: Optimal Transport-Guided Multi-Bridge Flow for Multimodal Sequential Recommendation (OT-FlowRec) — Nanjing University of Science and Technology, arXiv 2610.16038- DPE-Rec: Diversity-Preserving and Efficient Diffusion Language Model Recommendation for Bundle Construction (DPE-Rec) — Arizona State University, arXiv 2610.34700 ### Distributed Training / Systems - Efficient and Robust Online Learning to Rank in Decentralized Systems (RankGuard) - Building a privacy-preserving Federated Recommender system for mobile devices - Towards Generalizable and Efficient Large-Scale Generative Recommenders - FreeScale (Distributed Training for Sequence Recommendation) - RcLLM: Accelerating Generative Recommendation via Beyond-Prefix KV Caching - Scaling Graph Neural Networks for Friend Recommendation: Multi-Hash User Embeddings and Temporal Neighbor Sampling - Kernel-Managed Shared Memory for System-Wide Personalization - Optimizing Effective Training Time for Large-Scale Recommendation Systems — Meta, arXiv 2610.02057 - GroundedGEO: Auditing the Evidence Gap in Generative Search Rankings (GroundedGEO) — Shenzhen University, arXiv 2609.25189 ### Evaluation / Benchmark - The Recall Ceiling of LLM Recommendation Reranking (RecallCeiling) — University of Southern California, arXiv 2609.27953 - Bradley-Terry Rankings for Recommender Systems Across Dataset Taxonomies - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - Do Sequential Recommendation Benchmarks Really Require Higher-Order Sequence Modelling? (SeqBench) - From Prompt to Purchase: How AI Brand Recommendations Move Consumers on the Open Web - LLM-as-a-Judge for Reliable and Explainable Offline Evaluation in Top-K Recommendation (LLM Judge) - One Polluted Page Is Enough: Evaluating Web Content Pollution in Generative Recommenders (FORGE) - OneRetrieval: Unifying Multi-Branch E-commerce Retrieval with an Editable Generative Model (OneRetrieval) - OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework (OneSearch-V2) - Can We Steer the Black-Box? Controllability-Centric Evaluation of Recommender Systems (CtrlBench-Rec) - PaperFlow: Profiling, Recommending, and Adapting Across Daily Paper Streams (PaperFlow) - Popcorn: A Configurable Benchmark for Visual Evidence in Multimodal Movie Recommendation (Popcorn) - SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents (SafeGEO) - SocialPersona: Benchmarking Personalized Profiling and Response with Multimodal Social-Media Context (SocialPersona) - SIDInspector: A Mapping-First Diagnostic Resource for Semantic-ID Tokenizers - Statistically Reliable LLM-Based Ranking Evaluation via Prediction-Powered Inference (PRECISE) - tau-Rec: A Verifiable Benchmark for Agentic Recommender Systems - Towards Fast Domain Adaptation and Fine-Grained User Simulation for Evaluating Conversational Recommender Systems (AdaptSim) - Understanding and Debugging Failures in N-Gram-Based Generative Retrieval - Variance Reduction for Heavy-Tailed Monetization Metrics in Ranking Experiments via Post-Stratification (ShareChat/SIGIR 2026) - Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models (LLM Brand) - AgentSelect: Benchmark for Narrative Query-to-Agent Recommendation (AgentSelect) - Are We Really Making Progress in Group Recommendation? Unmasking the Tie-Breaking Illusion (Tie-Breaking) - Overview of the TREC 2025 Product Search and Recommendation Track - SWIM: Step-Wise Integrated Measure for Session-supervised List Evaluation in Generative Re-ranking (SWIM) - Equal Ranking Quality, Different Decisions: Training Order-Consistent LLM Scorers - When Does Supervised Fine-Tuning Reduce Instruction Sensitivity? - An Empirical Evaluation of Cross-City POI Recommendation on a Large-Scale Benchmark - Off-Policy Evaluation for Semantic ID Recommenders: Does the Model's Own Code Hierarchy Help? - RPCBench: A Benchmark for Proactive Premise Critique in LLM-based Recommendation (RPCBench) - The Utility of LLMs in Recommender Systems Explanation Evaluation - The Dice Roll Method: A Standardized Protocol for Repeated-Query Auditing of Large Language Model Brand Recommendations - Repeated Queries Exhaust an LLM's Brand Recommendations but Not Its Sources - FINALLY: A Dataset Recommender System for Recommender-Systems Research (FINALLY) - REDSI: Addressing the Reproducibility and Evaluation Consistency of Differentiable Search Indexing for Document Retrieval (REDSI) - PRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong Conversations (PRAGMA) - Should I Be Polite to My LLM Relevance Judge? Tone as a Severity Operating-Point Shift - Evaluating Brand Retrieval and Ranking in Large Language Model Recommendations - LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory, with ICE v2 as an Audited Local-First Architecture (LSREP) - RegRet: Enhancing Region-Level Retrieval in Large Multimodal Models (RegRet) - Understanding AI Provider Recommendations in Local Service Markets - Reproducing Transparent and Scrutable Recommendations: Exploring Open-Weight Models via Natural-Language User Profiles (Transparent UPR Repro) — University of Zurich, BlackBoxNLP @ EMNLP 2026 - Dense Feature Representation over Sequence Modeling: A Solution to the KDD Cup 2026 UniRec Challenge (KDD Cup 2026 UniRec) — Z Lab (Chengdu), KDD Cup 2026 UniRec Workshop - Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation (CRS-Performance) - The Lifecycle of LLM-as-a-Judge for Large-Scale Recommendation Explanations - From Prompting to Behavioral Alignment: Personalized LLM Judges for Recommendation Evaluation - RecToolBench: Benchmarking Recommendation-Specific Tool Orchestration under Fuzzy User Intent (RecToolBench) — University of Virginia / Jilin University / Squirrel AI / PolyU, arXiv 2609.30717 - ED-DR: Off-Policy Ranking Evaluation under Ranking-Dependent Examination (ED-DR) - Waseda University, arXiv 2609.35034 - Eval4DiRec: A Unified and Systematic Evaluation Framework for Diffusion-based Recommender Systems (Eval4DiRec) - University of Technology Sydney, ACM TKDD, arXiv 2609.34404 - A Behavioral Trait Leaks into Preferences: Diagnosing Trait Interference in LLM User Simulators (PQA) — KAIST, arXiv 2609.25572 [opensource] - When the Label Ignores the Request: Auditing Policy-Selected Targets in Synthetic Conversational Music Recommendation (When-the-Label) — Uber AI, arXiv 2609.39696 [opensource] - UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents (UserArena) — Leiden University / Renmin University of China / Alibaba Group / Institute of Computing Technology, CAS, arXiv 2610.30932 [opensource] ### Fairness - Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems — Pinterest, arXiv 2609.33073 - FARE: Deep Reinforcement Learning for Fair Exposure Constrained Uncertainty-Aware Financial Content Personalization (FARE) — JPMorganChase, arXiv 2609.31890 - A Rank-One Popularity Component in Dot-Product Recommender Scores: Population Theory and Prior-Separation Evidence - Adaptive Autoguidance for Item-Side Fairness in Diffusion Recommender Systems (A2G-DiffRec) - Attention Calibration for Position-Fair Dense Information Retrieval - Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation (CausalDPO) - Causal-Invariant Cross-Domain Out-of-Distribution Recommendation (CICDOR) - Causal Representation Learning for Generalisable Recommendation - Consensus vs. Dissent: Dynamic LLM Modeling of Subjective Preferences in Group Recommenders - Contextual Scalarisation Thompson Sampling for Multi-Objective Decisions in Public Media (CSTS) - COPF: An Online Framework for Deployment-Stable Counterfactual Fairness in Evolving Graphs - CRAB: Codebook Rebalancing for Bias Mitigation in Generative Recommendation - Distributional Approximate Nearest Neighbour Search for Uncertainty-Aware Retrieval (DINOSAUR) - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors - Echoes in Filter Bubble: Diagnosing and Curing Popularity Bias in Generative Recommenders (Ghost) - From Head to Tail: Asymmetric Knowledge Transfer in Long-tail Recommendation with Generative Semantic IDs - Gender and Race Bias in Consumer Product Recommendations by Large Language Models - Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems - LLM Biases - LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems - One Polluted Page Is Enough: Evaluating Web Content Pollution in Generative Recommenders (FORGE) - CoSimRec: Measuring Coordinated-Content Penetration in Recommender Feedback Loops (CoSimRec) - Recommendation Quality and the Concentration of Consumption: Experimental Evidence from Netflix - Representation Curriculum: Stagewise Training for Robust Ranking and Allocation - The Injection Paradox: Brand-Level Suppression in Safety-Trained LLM Recommendations via RAG Context Injection (Injection Paradox) - The Pitfall of Scaling Up: Uncovering and Mitigating Popularity Bias Amplification in Scaling Transformer-based Recommenders (SPRINT) - Toward User Preference Alignment in LLM Recommendation via Explicit Context Feedback - TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation (TRACER) - Trading Engagement for Sustainability: Carbon-Aware Re-ranking for E-commerce Recommendations - Trustworthy Recommendation in the Era of Large Language Models: Opportunities and Challenges - Uncertainty and Fairness Awareness in LLM-Based Recommendation Systems - When Recommendation Denoising Meets Popularity Bias: Understanding and Mitigating Their Interaction (PAD) - Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models (LLM Brand) - Whose Name Comes Up? III: Persona Prompting Effects in LLM-Based Scholar Recommendation - Designing Trustworthy LLM-based Wellbeing Recommendation through Controllable Interaction - From Hidden Profiles to Governable Personalization: Recommender Systems in the Age of LLM Agents - Scoring Is Not Enough: Addressing Gaps in Utility-fairness Trade-offs for Ranking - Rethinking Fairness in LLM-Based Recommender Systems: A Survey - SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents (SafeGEO) - Reproducing FACTER: Fairness via Conformal Thresholding and Prompt Repair - Fairness Attacks on Recommender Systems - Federated Variational Preference Alignment with Gumbel-Softmax Prior for Personalized User Preferences (FedVPA-GP) - PinEqualizer: Full Funnel Content Exploration and Debiasing System at Pinterest (PinEqualizer) - GRACE: Bringing GRACE to Recommendation: Fine-Tuning for Sustainable and Accurate Personalization (GRACE) - FSGR: Mitigating Token Frequency Bias for Fair SID-Based Generative Recommendation (FSGR) - Unbiased Recommender Systems with Implicit Feedback - Personalized and Multi-View Representation for Federated Cold-Start Recommendation (PMFRec) - SwapRec: Warming Up Cold Items Through Training-Time Swaps (SwapRec) - GenCAR: Generative Counterfactual Alignment with Risk-Controlled Selection for Out-of-Distribution Recommendation (GenCAR) - MGDiff: Multi-Interest Sequence Recommendation with Masking GNN-Guided Diffusion (MGDiff) - The Dice Roll Method: A Standardized Protocol for Repeated-Query Auditing of Large Language Model Brand Recommendations - Following the Preference, Missing the Optimum: Compliance Without Optimization in AI Housing Recommendation - What Price Fairness? Evaluating Energy - Fairness - Accuracy Trade-off in Recommender Systems - Quantifying the Engagement Trap: Impact of Short-form Video Recommender Systems on Users with ADHD - FunnelAudit: Responsibility Auditing in Multi-Route Recommender Systems (FunnelAudit) - Democracy Needs Reach: Political Equality, Online Speech, and Algorithmic Recommendation - Safety as a Constraint: Fine-Tuning a LLM Recommender to Explain Itself (Netflix / UPenn) - PCap: Personalized Retrieval-Stage Diversity Capping in Facebook Marketplace (PCap) - Evaluating Brand Retrieval and Ranking in Large Language Model Recommendations - Understanding AI Provider Recommendations in Local Service Markets - FacetCRS: Multi-Faceted Preference Learning for Pricking Filter Bubbles in Conversational Recommender System (FacetCRS) — Sun Yat-sen University, AAAI 2024 - Reproducing Transparent and Scrutable Recommendations: Exploring Open-Weight Models via Natural-Language User Profiles (Transparent UPR Repro) — University of Zurich, BlackBoxNLP @ EMNLP 2026 - Give the Long-tail More SPACE: Promoting Provider Fairness in Next POI Recommendation (SPACE) — Southeast University, arXiv 2608.07998 - Fair on the Surface? Benchmarking Hidden-Output Fairness Gaps in LLM Recommenders (FairGap) — University of Georgia / USC / CMU / Michigan State; arXiv 2608.08284 - Mult-BiW: Mitigating Popularity Bias via Multinomial-Likelihood Bi-Weighting (Mult-BiW) - Université de Montréal, ACM TOIS, arXiv 2609.35041 - Measuring and Mitigating Identity-Cue Preference Drift in LLM-based Recommender Systems (PromptShift) — UESTC (training-free bias mitigation) - FairDiff: Mitigating the Self-Reinforcing Matthew Effect in Diffusion Recommender Models (FairDiff) — Tsinghua University / Huawei Noah's Ark Lab / USTC, arXiv 2609.36671 - Generate What You Can Trust: Content Credibility in Generative Recommenders (CreGR) — University of Technology Sydney, arXiv 2610.05670 [opensource] - Aligning Performance with Contribution: Towards Contribution-Aware Fair Recommendation (CPFR) — Shanghai University of Finance and Economics / Singapore University of Technology and Design, arXiv 2610.08245 - SafeCRS: Personalized Safety Alignment for LLM-Based Conversational Recommender Systems (SafeCRS) — UIC / UIUC / Amazon, personalized safety alignment, KDD 2026 — arXiv 2603.03536 [opensource] - PR-entmax: Mitigating Popularity Bias in Sequential Recommendations with Sparse Prior-Regularized Objective (PR-entmax) — Sber AI Lab / Lomonosov MSU, CIKM 2026 — arXiv 2610.17823 - FAIRGate: Agentic Runtime Policy Selection and Verified Repair for Fairer Recommendation Delivery (FAIRGate) — University of Massachusetts Boston, arXiv 2610.30359 ### Feature Selection - LeAP: Learnable Adaptive Permutation for Feature Selection in Heterogeneous and Sparse Recommender Systems (LeAP) ### Federated Recommendation - Beyond Centralization: User-Controlled Federated Recommendations in Practice - Building a privacy-preserving Federated Recommender system for mobile devices - Federated User Behavior Modeling for Privacy-Preserving LLM Recommendation (SF-UBM) - Federated Variational Preference Alignment with Gumbel-Softmax Prior for Personalized User Preferences (FedVPA-GP) - Guiding Federated Graph Recommendation with LLM-encoded knowledge - FedCGR: Federated Cross-Domain Generative Recommendation (FedCGR) — CIKM 2026 - Personalized and Multi-View Representation for Federated Cold-Start Recommendation (PMFRec) - RegionFed: Federated Learning for Personalized Query Understanding in Heterogeneous Retail Environments (RegionFed) - FedHUR: Learning Hierarchical Utility-Guided Client Relations for Personalized Federated Recommendation (FedHUR) ### Foundation Models - Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation (Scaling Laws) - Scaling Laws for Behavioral Foundation Models over User Event Sequences - One Sequential Recommendation Model Pretrained from Synthetic Priors Predicts Multiple Datasets (SRPFN) - RecPFN: Prior-Fitted Networks for In-Context-Based Recommendations (RecPFN) - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - UniMixer: A Unified Architecture for Scaling Laws in Recommendation Systems (UniMixer) - IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation MoE (IntBMoE) — Alibaba (AMap); deployed in AMap generative rec, arXiv 2609.21346 - From a Static Multi-Level Small Semantic Codebook to a Dynamic Single-Level Large Semantic Codebook for Generative Recommendation — Kuaishou, arXiv 2608.21012 ### Generative Recommendation / VAE - TransX Technical Report: Scaling Sequence Transduction for Recommendation (TransX) — LinkedIn, arXiv 2610.20913 - Multi-Decoder OneRec: Controllable Generative Retrieval for Multi-Objective Industrial Recommendation (Multi-Decoder OneRec) - Once Generated, Ranked: End-to-End Generative Slate Recommendation with Unified Semantic-Collaborative IDs (OGR) - WhisperRec: Latent Reasoning for Efficient Foundation Recommendation Models (WhisperRec) - A Rank-One Popularity Component in Dot-Product Recommender Scores: Population Theory and Prior-Separation Evidence - A2Gen (Action-Aware Generative Sequence Modeling) - APAO: Bridging the Training-Inference Gap in Generative Recommendation via Adaptive Prefix-Aware Optimization (APAO) - Adaptive Loss Balancing for Noise-Robust GRPO in Generative Recommendation (AdaGRPO) - Affective Music Recommendation: A Rollout-Based World Model for Offline Preference Optimization (AMRS) - AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems (AgentX) - Asymmetric Generative Recommendation via Multi-Expert Projection (AsymRec) - Beyond Instance-Level Alignment and Uniformity: Semantic Factor Learning for Collaborative Filtering (SaFeAU) - Beyond Static Collision Handling: Adaptive Semantic ID Learning (AdaSID) - Beyond Dense Connectivity: Explicit Sparsity for Scalable Recommendation (SSR) - Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling - BITRec (Modeling Behavioral Intensity and Transitions) - CARD: Non-Uniform Quantization of Visual Semantic Unit for Generative Recommendation - Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation (CausalDPO) - Causal Representation Learning for Generalisable Recommendation - ChronoID: Infusing Explicit Temporal Signals into Semantic IDs for Generative Recommendation (ChronoID) - CRAB: Codebook Rebalancing for Bias Mitigation in Generative Recommendation - Deep Interest Mining (SemanticID) - DeGRe: Dense-supervised Generative Reranking for Recommendation - DiffCold: A Diffusion-based Generative Model for Cold-Start Item Recommendation (DiffCold) - Differentiable Semantic ID for Generative Recommendation (DIGER) - Diffusion-GR2: Diffusion Generative Reasoning Re-ranker (Diffusion-GR2) - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - Don't Let Bandit Feedback Pull Continual LLM-Recommender Updates Off Target (ABPO) - DREAM: Dynamic Refinement of Early Assignment Mappings (DREAM) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - Echoes in Filter Bubble: Diagnosing and Curing Popularity Bias in Generative Recommenders (Ghost) - EvoRec: Self Evolving Agentic Recommender Systems (EvoRec) - Expressiveness Limits of Autoregressive Semantic ID Generation in Generative Recommendation (Latte) - F-GRPO: Factorized Group-Relative Policy Optimization - FAVE: Flow-based Average Velocity Establishment for Sequential Recommendation - Feedback-Grounded Policy Discovery for Generative Recommendation (Feedback-Grounded) - Federated User Behavior Modeling for Privacy-Preserving LLM Recommendation (SF-UBM) - Filling the Gaps: Selective Knowledge Augmentation for LLM Recommenders (KnowSA_CKP) - From Bootstrapping to Sequence Modeling: A Unified Generative Framework for Personalized Landing-Page Modeling (GLAN) - From Head to Tail: Asymmetric Knowledge Transfer in Long-tail Recommendation with Generative Semantic IDs - From Noise to Order: Learning to Rank via Denoising Diffusion (DiffusionRank) - From "Strings" to "Things" for Personal Knowledge Graphs (PKG-Rec) - FusID: Modality-Fused Semantic IDs for Generative Music Recommendation (FusID) - GCIB: Graph Contrastive Information Bottleneck for Multi-Behavior Recommendation - Gated Bidirectional Linear Attention for Generative Retrieval (GBLA) - GEMs: Breaking the Long-Sequence Barrier in Generative Recommendation with a Multi-Stream Decoder (GEMs) - Generative Recommendation for Large-Scale Advertising (GR4AD) - Generative Sequential Recommendation via Hierarchical Behavior Modeling (GAMER) - GenPage: Towards End-to-End Generative Homepage Construction at Netflix (GenPage) - GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation - GPlan: Generative Spatiotemporal Intent Sequence Recommendation (GPlan) - GR2 Technical Report (GR2) - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - Grounded Token Initialization for New Vocabulary in LMs for Generative Recommendation (GTI) - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - A General Framework for Multimodal LLM-Based Multimedia Understanding in Large-Scale Recommendation Systems (Multimodal LLM Rec) - HoloRec: Holistic Encoding and Interleaved Reasoning for Generative Recommendation (HoloRec) - How Reliable Are Semantic-ID Tokenizer Comparisons in Generative Recommendation? - How Well Does Generative Recommendation Generalize? - HRPO: Hierarchical Residual Policy Optimization for Generative Recommendations (HRPO) - HSUGA: LLM-Enhanced Recommendation with Hierarchical Semantic Understanding - Hyperbolic RQ-VAE enhanced Generative Recommendation with Differential-Length Codebook Strategy (HG-Rec) - IAT: Instance-As-Token Compression for Historical User Sequence Modeling (IAT) - Implicit Reasoning for Large Language Model-based Generative Recommendation (PauseRec) - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation - Interests Burn-down Diffusion Process for Personalized Collaborative Filtering (StageCF) - Intuition-Guided Latent Reasoning for LLM-Based Recommendation (IntuRec) - LASAR: Latent Adaptive Semantic Aligned Reasoning for Generative Recommendation - Learning Variable-Length Tokenization for Generative Recommendation (VarLenRec) - LLM-as-a-Judge for Reliable and Explainable Offline Evaluation in Top-K Recommendation (LLM Judge) - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - LoopMemGR: From Behavior Logs to Evolving Memory for Generative Recommendation (LoopMemGR) - LWGR: Lagrangian-Constrained Personalized World Knowledge for Generative Recommendation - Memento: Personalized RAG-Style Long-Retention Data Scaling for META Ads Recommendation - MTServe: Efficient Serving for Generative Recommendation Models - Multi-Business Prediction for Generative Recommendation at Meituan (MBGR) - Music Playlist Captioning at Scale with Large Language Models - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - OMEGA: Collaborative Memory Augmentation for Generative Recommendation (OMEGA) - On the Memorization Behavior of LLMs in Generative Recommendation: Observations, Implications, and Training Strategies (IIRG) - One Model, Two Markets: Bid-Aware Generative Recommendation (GEM-Rec) - One Pool, Two Caches: Adaptive HBM Partitioning for Accelerating Generative Recommender Serving - OneBar: An End-to-End Content-Grounded Generative Query Recommendation Framework for E-Commerce Video Feeds (OneBar) - OneFeed: A Unified Generative Framework for Feed Content Enhancement and Query Generation (OneFeed) - OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation (OneLive) - OneReason Technical Report - Position-Aware Drafting for Inference Acceleration in LLM-Based Generative List-Wise Recommendation (PAD-Rec) - Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation (Scaling Laws) - PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation - Query-Conditioned Generative Search with QGS in Quark - R3-VAE - RAGR: Review-Augmented Generative Recommendation - RankGR: Rank-Enhanced Generative Retrieval with Listwise Direct Preference Optimization in Recommendation (RankGR) - Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models - RSIR: Can Recommender Systems Teach Themselves? A Recursive Self-Improving Framework with Fidelity Control (RSIR) - ReCast - Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language (Restoring Collaborative) - RRCM: Ranking-Driven Retrieval over Collaborative and Meta Memories for LLM Recommendation - SAERec: Constructing Fine-grained Interpretable Intents Priors via Sparse Autoencoders for Recommendation (SAERec) - SAPO: Step-Aligned Policy Optimization for Reasoning-Based Generative Recommendation - Semantic Trimming and Auxiliary Multi-step Prediction for Generative Recommendation (STAMP) - SIREN: Unified Multi-Granularity Semantic Interaction for Multi-Modal Lifelong User Interest Modeling - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - SimGR: Escaping the Pitfalls of Generative Decoding in LLM-based Recommendation (SimGR) - SPARC: Sequence-aware Progressive Attribute Routing and Compression Framework for Generative Recommendation (SPARC) - Stop Treating Collisions Equally: Qualification-Aware Semantic ID Learning for Recommendation at Industrial Scale (QuaSID) - Structuring and Tokenizing Distributed User Interest Context for Generative Recommendation (G2Rec) - SynGR: Unleashing the Potential of Cross-Modal Synergy for Generative Recommendation - TCA4Rec: Token-level Collaborative Alignment for LLM-based Generative Recommendation - Tencent Advertising Algorithm Challenge 2025: All-Modality Generative Recommendation - The Injection Paradox: Brand-Level Suppression in Safety-Trained LLM Recommendations via RAG Context Injection (Injection Paradox) - The Pitfall of Scaling Up: Uncovering and Mitigating Popularity Bias Amplification in Scaling Transformer-based Recommenders (SPRINT) - Time-Aware Diffusion based on Preference Disentanglement for Generative Recommendation (TDPM) - TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems (TokenMinds) - Towards Fast Domain Adaptation and Fine-Grained User Simulation for Evaluating Conversational Recommender Systems (AdaptSim) - Towards Generalizable and Efficient Large-Scale Generative Recommenders - Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search (GrowthGR) - TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation (TRACER) - TriAlignGR: Triangular Multitask Alignment with Multimodal Deep Interest Mining for Generative Recommendation - UFRec: Uncertainty-Guided Future Learning for Sequential Recommendation - Unified Value Alignment for Generative Recommendation in Industrial Advertising (UniVA) - UniRec: Bridging the Expressive Gap via Chain-of-Attribute - URecJPQ: Memory-efficient Multimodal Recommendation Models through RecJPQ in Large-Scale Scenarios (URecJPQ) - When Text-as-Vision Meets Semantic IDs in Generative Recommendation: An Empirical Study - Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models (LLM Brand) - Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale (RaG) - The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation (H2Rec) - TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems (TokenMinds) - TRUST: Item-Calibrated Interval Evidence for Temporal Session-Based Recommendation (TRUST) - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - UniMixer: A Unified Architecture for Scaling Laws in Recommendation Systems (UniMixer) - LLM-HYPER: Generative CTR Modeling for Cold-Start Ad Personalization via LLM-Based Hypernetworks (LLM-HYPER) - Next-Scale Generative Reranking: A Tree-based Generative Rerank Method at Meituan (NSGR) - MLPs are Efficient Distilled Generative Recommenders (SID-MLP) - TwiSTAR: Think Fast, Think Slow, Then Act — Generative Recommendation with Adaptive Reasoning (TwiSTAR) - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation (Tencent/ACL 2026) - Interpretable Representation via LLM-Driven Generative Disentanglement for Local-Life Service Recommendation (LGRID) - HCGRec: Hint-Conditioned Generative Recommendation with Semantic IDs (HCGRec) - Sona Technical Report (Sona) - Gwhere: Guess Where You Go — Generative Next Point-of-Interest Recommendation in Amap (Gwhere) - EchoRec: Multi-Item Prediction-Empowered Generative Recommendation via Cycle-Consistent Preference Alignment (EchoRec) - Decoupled Temporal Encoding for Generative Recommendation (DTE) - SAGA: Structure-Attended Generative Action Embedding Model that encodes Multi-Surface User Action Sequences (SAGA) - ReliGRec: Reliability-Oriented LLM-Based Generative Recommendation via User-Risk-Aware Prompt Routing (ReliGRec) - LIGE-GR: A Smooth Leap from Ranking to Generative Recommendation in the LLM Era (LIGE-GR) - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 - PAPA: Online Personalized Active Preference Alignment (PAPA) — Washington University in St. Louis, arXiv 2607.00486 - Iterative Semantic Reasoning from Individual to Group Interests for Generative Recommendation with LLMs (ISRF) - Beyond Interleaving: Causal Attention Reformulations for Generative Recommender Systems - GroundedGEO: Auditing the Evidence Gap in Generative Search Rankings (GroundedGEO) — Shenzhen University, arXiv 2609.25189 - ReFilter: Bridging Embeddings and LLM Filtering for Similar Mobile App Retrieval (ReFilter) — University of Toronto / Université du Québec à Montréal; ASIS&T 2026, arXiv 2609.25306 - T-RoPE: Time-Aware Rotary Position Embedding for Sequential Recommendation (T-RoPE) — Shopify, arXiv 2609.30576 - KuaFu: Compressing Long User Behavior into Understanding at Billion Scale (KuaFu) — Tencent, arXiv 2609.31045 - Recommendation World Models for Future-State Control (UA-TWM) — University of British Columbia et al., arXiv 2609.30711 - GRP: Snap's Generative Recommendation Paradigm - Unified Retrieval, Ranking, and Reward Modeling (GRP) - Snap Inc., arXiv 2609.36688 - Generate What You Can Trust: Content Credibility in Generative Recommenders (CreGR) — University of Technology Sydney, arXiv 2610.05670 [opensource] - Adapting Generative Recommenders for Multi-Turn Interaction (INTEGER) — National Taiwan University / Johns Hopkins University, arXiv 2610.08136 - HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation (HASTE) — Glance, arXiv 2610.17368 - TacGR (UniSGR-V2): Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation (TacGR) — Alibaba International, arXiv 2610.16132 - Structure-Aligned History Learning with Subspace-Structured Semantic IDs for Generative Recommendation (SAHL) — Chongqing University, arXiv 2610.25093 - OneChord: Distill What You Rank, Recommend in One Pass (OneChord) — University of Chinese Academy of Sciences, arXiv 2610.25805 - From Scores to Safe Sets: Exact Safety-Constrained Decoding for Medication Recommendation (SafeSets) — Optum, arXiv 2610.29228 ### Generative Retrieval / Ranking - ReGen: Controllable Generative Recommendation via Guided Token Refinement (ReGen) — XJTLU / U Liverpool / PolyU, arXiv 2610.28642 [opensource] - Generative Recommenders Forget the Long Tail: Restoring Memorization in Retrieval and Ranking (CAPE) — USTC & Ant Group, arXiv 2610.29298 - Evolving Semantic Indexes for Generative Retrieval — Santa Clara University & Coupang, arXiv 2610.28568 - Revisiting Generative vs. Dense Retrieval: A Static Value Prior Reverses the Gap (Static Value Prior) — Shopify, arXiv 2610.19065 - OneTrans-V2: Unifying Retrieval, Pre-rank, and Fine-rank with One Transformer in Industrial Recommender (OneTrans-V2) — ByteDance, arXiv 2609.28589 - X-Rec Technical Report (X-Rec) — TikTok, arXiv 2609.29180 - Beyond Matching: Category-Guided Latent Intent Reasoning for Generative Retrieval in E-Commerce (CaLIR) - CapsID: Soft-Routed Variable-Length Semantic IDs - CARD: Non-Uniform Quantization of Visual Semantic Unit for Generative Recommendation - Causal Representation Learning for Generalisable Recommendation - ChronoID: Infusing Explicit Temporal Signals into Semantic IDs for Generative Recommendation (ChronoID) - Closing the Indexing-Decoding Gap in Multimodal Generative Retrieval via Prefix Retention Optimization (PRO) - DIRECTOR: Dynamic Index-based Recommendation with Transport-Optimized Retrieval (DIRECTOR) - Multi-Decoder OneRec: Controllable Generative Retrieval for Multi-Objective Industrial Recommendation (Multi-Decoder OneRec) - PSG: Pair-Space Generation for Efficient Generative Reranking (PSG) - DaV-Gen: End-to-End Generative Retrieval via Draft-and-Verify (DaV-Gen) - Decoupled Residual Quantization for Robust Semantic IDs in Recommendation (DRQ) - DeGRe: Dense-supervised Generative Reranking for Recommendation - Differentiable Geometric Indexing for End-to-End Generative Retrieval (DGI) - Diffusion-GR2: Diffusion Generative Reasoning Re-ranker (Diffusion-GR2) - Discrimination Is Generation: Unifying Ranking and Retrieval from a Tokenizer Perspective (DIG) - DOS: Dual-Flow Orthogonal Semantic IDs for Recommendation in Meituan (DOS) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - EviRank: Evidence-Based Confidence Estimation for LLM-Based Ranking (EviRank) - Expressiveness Limits of Autoregressive Semantic ID Generation in Generative Recommendation (Latte) - Feedback-Grounded Policy Discovery for Generative Recommendation (Feedback-Grounded) - Next-Scale Generative Reranking: A Tree-based Generative Rerank Method at Meituan (NSGR) - From Noise to Order: Learning to Rank via Denoising Diffusion (DiffusionRank) - Gated Bidirectional Linear Attention for Generative Retrieval (GBLA) - Generative Reasoning Re-Ranker (GR2) - Generative Retrieval via Diffusion Transformer with Metric-Ordered Sequence Training and Hybrid-Policy Preference Optimization (MO-DiT+HPPO) - Generative Sequential Recommendation via Hierarchical Behavior Modeling (GAMER) - GenPage: Towards End-to-End Generative Homepage Construction at Netflix (GenPage) - GLIDE: Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify (GLIDE) - SmartGR: Hierarchy and Beam-Aware Knowledge Distillation for Generative Recommendation (SmartGR) - SSR-GRPO: Integrating Supervision and Semantic IDs into Reinforcement Learning for Dense Retrieval in E-commerce (SSR-GRPO) - UniR²: Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence (UniR²) - GPlan: Generative Spatiotemporal Intent Sequence Recommendation (GPlan) - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - Harmonizing Generative Retrieval and Ranking in Chain-of-Recommendation (RecoChain) - IID-Nav: From Extraction to Navigation: Progressive Retrieval with Indirectly Infinite Depth (IID-Nav) - LLMs Need Encoders for Semantic IDs Too (PrefixMem) - LRanker: LLM Ranker for Massive Candidates - Let the Agent Steer: Closed-Loop Ranking Optimization via Influence Exchange (Sortify) - MERGE: Next-Generation Item Indexing Paradigm for Large-Scale Streaming Recommendation (MERGE) - Mitigating Collaborative Semantic ID Staleness in Generative Retrieval - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - OneBar: An End-to-End Content-Grounded Generative Query Recommendation Framework for E-Commerce Video Feeds (OneBar) - OneFeed: A Unified Generative Framework for Feed Content Enhancement and Query Generation (OneFeed) - OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation (OneLive) - OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System (OnePiece) - OneRetrieval: Unifying Multi-Branch E-commerce Retrieval with an Editable Generative Model (OneRetrieval) - OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework (OneSearch-V2) - PIT: A Dynamic Personalized Item Tokenizer for End-to-End Generative Recommendation (PIT) - Prompt Optimization for User Simulation in Conversational Recommender Systems (UserSimulator) - PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation - Quantizing Intent: Cross-Domain Semantic IDs from Organic Activity for Industrial Ranking - Query-Conditioned Generative Search with QGS in Quark - RankGR: Rank-Enhanced Generative Retrieval with Listwise Direct Preference Optimization in Recommendation (RankGR) - Rethinking Sales Lead Scoring with LLM-based Hierarchical Preference Ranking (HPRO) - Revisiting General Map Search via Generative Point-of-Interest Retrieval (GenPOI) - RGD: Reward Guided Decoding for Generative Recommendation (RGD) - SSRLive: Live Streaming Recommendation with Dynamic Semantic ID (SSRLive) - ThinkGR: Integrating Chain-of-Thought into Generative Retrieval - TriAlignGR: Triangular Multitask Alignment with Multimodal Deep Interest Mining for Generative Recommendation - Understanding and Debugging Failures in N-Gram-Based Generative Retrieval - UniNote: A Unified Embedding Model for Multimodal Representation and Ranking - UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale - Planning over Matrix-Factorization MDPs for Candidate Generation (MF-MDP Planning) - Prompt Generation Technical Report (PG) - Beyond Semantic IDs: Encoding Business-Value Ranking into Document Identifiers for Generative Retrieval (CRID) - Token-Level Credit Assignment Optimization for Generative Document Retrieval - Generative Universal Multimodal Retrieval with Dual-role Identifiers (DrIG) - Improving Item Discoverability in e-Commerce Search via Related Intent Generation - Hypothesis-Driven Shelf Generation for Personalised Recommendation - ICICLE: Expanding Retrieval with In-Context Documents (ICICLE) - SWIM: Step-Wise Integrated Measure for Session-supervised List Evaluation in Generative Re-ranking (SWIM) - PailitaoGR: Latent Think-with-Images for Generative Image Retrieval (PailitaoGR) - ProRetrieval: Learning to Orchestrate Hybrid Search via Executable Program Synthesis (ProRetrieval) - Preference Shapes Relevance: Cross-component Hierarchical Semantic Alignment for Personalized Generative Retrieval (CHAP) - Generative Retrieval for E-commerce: Jointly Learning Embedding and Codebook with Same Product Cluster - TGR: Tencent Generative Recommendation — Unified Generation and Reasoning (TGR) - It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning (CoGR) - SPAR: Enhancing Industrial-Scale Generative POI Recommendation via Real-World Spatial Perception (SPAR) - ICEGR: An Intent-Coherent End-to-End Generative Retrieval Framework for E-commerce Search (ICEGR) - HF-SID: High-Fidelity Semantic IDs for Generative Retrieval in Location-Based Services (HF-SID) - SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce (SAM-D2Q) - SAGE: Semantic Attribute Graphs for Multi-Entity Visual Retrieval (SAGE) - Inventory-Grounded Policy-Level Optimization for Training-Free AI Search (IGPO) - PDMR: Passage-Driven Multi-ID Document Retrieval (PDMR) - REDSI: Addressing the Reproducibility and Evaluation Consistency of Differentiable Search Indexing for Document Retrieval (REDSI) - Exploring Bottom-Up Clustering for Creating Semantic IDs - Recommendation Retrievers Need Verifiers: Universal Generative Reranking for Sequential Recommendations (Verifiers) - VARG: Value-Aware and Ranking-Aligned Generative Retrieval for Dynamic E-commerce Search (VARG) - One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations (ANGLE) - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 - SELF-INDEX: Self-Evolving Search Index (SELF-INDEX) — Yonsei University, arXiv 2609.19656 - GRP: Snap's Generative Recommendation Paradigm with Multimodal Semantic IDs (GRP) - Snap Inc., arXiv 2609.36688 - SpeakGR: Learning Semantic IDs Without Forgetting Language Generation (SpeakGR) - Imperial College London, arXiv 2609.35430 - Beyond the Beam: Constructive Repair and Candidate Completion for Generative Recommendation (Beyond the Beam) — Beijing Institute of Technology - RankEvolve: A Reliable Multi-Agent Auto-Research Harness for Evolving Ranking Models (RankEvolve) — Meta, arXiv 2609.39551 - SPRINT: Single-Step Generative Recommendation via Average Probability Velocity (SPRINT) — University of Technology Sydney (single-pass SID generation) - SPRIG: Semantic-ID-enhanced Paths for Knowledge Graph-based Generative Recommendation (SPRIG) — Johannes Kepler University Linz, CIKM 2026, arXiv 2610.06590 [opensource] - A Systematic Study of Semantic ID Spaces for Generative Information Retrieval (Semantic-ID Spaces) — Artefact Research Center / Université d'Angers (LERIA), arXiv 2610.08732 - Disentangling Paradigm, Identifier, and Decoding in Generative Retrieval (Disentangling Paradigm/Identifier/Decoding) — Artefact Research Center / Université d'Angers (LERIA), arXiv 2610.08716 - Missed Targets: Training with Missed Targets in Generative Recommendation (Missed Targets) — Zhejiang University, arXiv 2610.10124 - LIFT: Lifecycle-Aware Interaction Factorization for Unified Retrieval and Ranking (LIFT) - Beijing University of Posts and Telecommunications / Southeast University / Xiaohongshu, arXiv 2610.10556 - Bit-Factorized: Bit-Factorized Identifiers for Efficient Generative Retrieval (Bit-Factorized) — Leiden University / Radboud University, arXiv 2610.17467 - WeGR: Unified Generative Retrieval for Industrial Search Systems in WeChat (WeGR) — Tencent (WeChat), arXiv 2610.16979 - AGR: Agentic Generative Retrieval with Tool-Augmented Search (AGR) — Kuaishou Technology & Fudan University, arXiv 2610.26831 - Don't Rewrite the SID Map: Topology Preservation for Generative Retrieval in Dynamic E-Commerce Search (SID-Map) — Alibaba, arXiv 2610.26105 ### Graph-based Recommendation - Adapting Knowledge Graphs for Behavior Denoising in Sequential Recommendation (AdaptedKG) - Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation (Shallow-RHS) - COPF: An Online Framework for Deployment-Stable Counterfactual Fairness in Evolving Graphs - GCIB: Graph Contrastive Information Bottleneck for Multi-Behavior Recommendation - Macro Graph of Experts for Billion-Scale Multi-Task Recommendation (MGOE) - Generalizing Graph Foundation Models via Hyperbolic Retrieval-Augmented Generation (HyRAG) - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - GraphRAG-IRL - Guiding Federated Graph Recommendation with LLM-encoded knowledge - POI Recommendation with LLM-Augmented Multi-Graph Learning and Contrastive Alignment (LLM-MGCL) - RankGraph-2: Lifecycle Co-Design for Billion-Node Graph Learning in Recommendation (RankGraph-2) - Structuring and Tokenizing Distributed User Interest Context for Generative Recommendation (G2Rec) - CoPersona: Collaborative Persona Graphs for Robust LLM Personalization (CoPersona) - CCLRec: Consensus-driven Contrastive Learning for LLM-enhanced Graph Recommendation (CCLRec) - Making Collaborative Signals Count: Graph-Aware Large Language Models for Sequential Recommendation (GALLM) - Scaling Graph Neural Networks for Friend Recommendation: Multi-Hash User Embeddings and Temporal Neighbor Sampling - Continual Graph Memory for Adaptive Recommendation under Intent Drift (CGM-Rec) - MURAL: Multimodal Uncertainty-aware Recommendation via Adaptive edge Learning (MURAL) - Do All Nodes Benefit Equally from Knowledge Graphs? Adaptive Node-Aware KG Fusion for Recommendation (AdaKG) - PCGNet: Unifying Shared and Specific Information for Fashion Matching Recommendations (PCGNet) — The Hong Kong Polytechnic University, arXiv 2609.13339 - X-KGRank: A Knowledge Graph RAG Framework for Explainable Recommendations via Pattern Mining and LLM Re-Ranking (X-KGRank) — San Jose State University; arXiv 2608.01732 - GEAR: Generative End-to-end Ad Retrieval at Douyin (GEAR) — ByteDance (Douyin); arXiv 2609.39327 - Residual Trajectory Distillation for Generative Retrieval (ResTD) — Beihang University / Meituan; arXiv 2609.39319 - Learning Multiresolution Relevance for Hierarchical Generative Retrieval (RARS) — Beihang University / Meituan; arXiv 2609.39312 [opensource] - Exploring Forum Post Retrieval with Generative Modeling — Meta (Facebook); arXiv 2609.38646 - Graph-Informed Semantic IDs (GrIS): Balancing Semantic and Collaborative Signals via Recursive Graph Partition (GrIS) — Huawei Ireland Research Centre, arXiv 2610.01533 - SPRIG: Semantic-ID-enhanced Paths for Knowledge Graph-based Generative Recommendation (SPRIG) — Johannes Kepler University Linz, CIKM 2026, arXiv 2610.06590 [opensource] - Multimodal Graph Retrieval-Augmented Sequential Recommendation via Collaborative Filtering Paths (MGRASRec) — University of New South Wales, arXiv 2610.11228 ### Group Recommendation - AgentGR: Semantic-aware Agentic Group Decision-Making Simulator for Group Recommendation - Consensus vs. Dissent: Dynamic LLM Modeling of Subjective Preferences in Group Recommenders - Are We Really Making Progress in Group Recommendation? Unmasking the Tie-Breaking Illusion (Tie-Breaking) - Enhancing Group Recommendation with Memory-Augmented Reasoning in LLM Agent (AGR) - Iterative Semantic Reasoning from Individual to Group Interests for Generative Recommendation with LLMs (ISRF) - IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation MoE (IntBMoE) — Alibaba (AMap); deployed in AMap generative rec, arXiv 2609.21346 ### Industrial - RecSys Factory: Bounding LLM Agent Autonomy to Decision Points in the Industrial Recommender Lifecycle (RecSys Factory) — Tencent, arXiv 2608.11241 - EASformer: Environment-Action-State Alignment for Industrial Generative Ranking (EASformer) — Beihang University, arXiv 2610.27691 - TransX Technical Report: Scaling Sequence Transduction for Recommendation (TransX) — LinkedIn, arXiv 2610.20913 - Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops — Spotify, arXiv 2609.30297 - On Evaluating and Improving Conversational Agents in Production — Zalando, arXiv 2609.32092 - OneTrans-V2: Unifying Retrieval, Pre-rank, and Fine-rank with One Transformer in Industrial Recommender (OneTrans-V2) — ByteDance, arXiv 2609.28589 - X-Rec Technical Report (X-Rec) — TikTok, arXiv 2609.29180 - Advancing Model Research in AgentX: Long-Horizon Autonomy for Industrial Recommender Systems (AgentX-Model) — Kuaishou, arXiv 2609.30001 - Cross-Country Code-Mixing for Generative Recommendation (CMRec) - Decomposing Staleness in Recommender Systems: A Dual-Filter Framework for Supersession and Decay (SDF) - Discriminative Language Models as Effective Retrievers (Discriminative Retrieval) - DIRECTOR: Dynamic Index-based Recommendation with Transport-Optimized Retrieval (DIRECTOR) - EGR: Embedding-Native Generative Retrieval with a Shared LLM (EGR) - GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling (GateDiffInt) - Grevo: A Unified Generative Recommendation Framework with Evolutionary Item Indexing (Grevo) - Gryphon-v2: One Model in Place of a Cascade — Generate-and-Rank Recommender with Rollout Distillation (Gryphon-v2) - Multi-Decoder OneRec: Controllable Generative Retrieval for Multi-Objective Industrial Recommendation (Multi-Decoder OneRec) - Once Generated, Ranked: End-to-End Generative Slate Recommendation with Unified Semantic-Collaborative IDs (OGR) - OneModel: A Unified Foundation for Platform-Scale Multi-Scenario Ranking (OneModel) - OxygenREC-v2: Internalizing Discrimination into Generative Recommendation (OxygenREC-v2) - PSG: Pair-Space Generation for Efficient Generative Reranking (PSG) - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - SITA: Semantic Interest Tokens for Target-Aware Compression in Long-Sequence Recommendation (SITA) - SSR-GRPO: Integrating Supervision and Semantic IDs into Reinforcement Learning for Dense Retrieval in E-commerce (SSR-GRPO) - STEPS: A Self-Triggered Agentic Push Recommendation System (STEPS) - UniDot: A Unified Network for Sequence Modeling and Feature Interaction in Large-scale Recommendation (UniDot) - UniR²: Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence (UniR²) - WhisperRec: Latent Reasoning for Efficient Foundation Recommendation Models (WhisperRec) - LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation (LaRec) - OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation (OneLive) - PinRec: Unified Generative Retrieval for Pinterest Recommender Systems (PinRec) - SA²CRQ: Towards Efficient and Generalizable Retrieval — Adaptive Semantic Quantization and Residual Knowledge Transfer (SA²CRQ) - UniGD: A Unified Generative-Discriminative Framework for Industrial Retrieval (UniGD) - A General Framework for Multimodal LLM-Based Multimedia Understanding in Large-Scale Recommendation Systems (Multimodal LLM Rec) - Adaptive Re-Ranking - Affective Music Recommendation: A Rollout-Based World Model for Offline Preference Optimization (AMRS) - AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems (AgentX) - Atomic Intent Reasoning: Bringing LLM Semantics to Industrial Cross-Domain Recommendations (AIR) - BAHSD: Bridging the Long-tail Gap via Adaptive Distillation in Black-box Sequential Recommendation (BAHSD) - Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling - Bridging Short Videos and Live Streams: Reasoning-Guided Multimodal LLMs for Cross-Domain Representation Learning (RGCD-Rep) - Causal Representation Learning for Generalisable Recommendation - Cheaper is Better: A Discount-Aware Network for Conversion Rate Prediction in E-commerce Recommendation System (DANet) - CMSL: Constructive Multi-Sequence Learning for Recommendation Systems (CMSL) - Contextual Scalarisation Thompson Sampling for Multi-Objective Decisions in Public Media (CSTS) - Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking (CA-PG) - Decoupled Residual Quantization for Robust Semantic IDs in Recommendation (DRQ) - DIG: Discrimination Is Generation — Unifying Ranking and Retrieval from a Tokenizer Perspective (DIG) - DeGRe: Dense-supervised Generative Reranking for Recommendation - Denoising Implicit Feedback for Cold-start Recommendation (DIF) - DeRes: Decoupling Residual Stability and Adaptivity for Scalable CTR Prediction (DeRes) - DOS: Dual-Flow Orthogonal Semantic IDs for Recommendation in Meituan (DOS) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - Dual-Stream MLP is All You Need for CTR Prediction (DS-MLP) - EvoRec: Self Evolving Agentic Recommender Systems (EvoRec) - Exp-RSFT: Exponential Reward Weighting for Fine-Tuning Generative Recommenders under Sparse and Noisy Feedback (Exp-RSFT) - GEMs: Breaking the Long-Sequence Barrier in Generative Recommendation with a Multi-Stream Decoder (GEMs) - GenPage: Towards End-to-End Generative Homepage Construction at Netflix (GenPage) - GLIDE: Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify (GLIDE) - GPlan: Generative Spatiotemporal Intent Sequence Recommendation (GPlan) - GRACE: Generative Recommender Acceleration Engine for Real-Time Ads Retrieval (GRACE) - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - Grevo: A Unified Generative Recommendation Framework with Evolutionary Item Indexing (Grevo) - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - GR2 Technical Report (GR2) - HiLaR: Hierarchical Latent Reasoning for LLM-based Recommendation (HiLaR) - IAT: Instance-As-Token Compression for Historical User Sequence Modeling (IAT) - IID-Nav: From Extraction to Navigation: Progressive Retrieval with Indirectly Infinite Depth (IID-Nav) - LLM-HYPER: Generative CTR Modeling for Cold-Start Ad Personalization via LLM-Based Hypernetworks (LLM-HYPER) - LLM-Based Re-Ranking for Real Estate Search - Long-term User Engagement Optimization through Model-agnostic Downstream Rewards Learning - Causal Retrieval Optimization for Efficient e-commerce Distribution (Pinterest) - SIF: Sample Is Feature — Beyond Item-Level, Toward Sample-Level Tokens for Unified Large Recommender Models (SIF) - LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction (LoopCTR) - Next-Scale Generative Reranking: A Tree-based Generative Rerank Method at Meituan (NSGR) - Let the Agent Steer: Closed-Loop Ranking Optimization via Influence Exchange (Sortify) - INSPIRE: Intent-aware Neural Sponsored Product Retrieval for E-commerce - Joint Optimization of Relevance and Engagement in Multi-Task Ranking for E-Commerce with Efficient LLM Supervision - L2Rec: Towards Dual-View Understanding of LLMs for Personalized Recommendation - LeAP: Learnable Adaptive Permutation for Feature Selection in Heterogeneous and Sparse Recommender Systems (LeAP) - LLM Retrieval for Stable and Predictable Ad Recommendations - LLM-Based User Personas for Recommendations at Scale - Macro Graph of Experts for Billion-Scale Multi-Task Recommendation (MGOE) - LLMs Need Encoders for Semantic IDs Too (PrefixMem) - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - LoopMemGR: From Behavior Logs to Evolving Memory for Generative Recommendation (LoopMemGR) - MDGR: A Masked Diffusion Generative Recommendation Framework (MDGR) - Memento: Personalized RAG-Style Long-Retention Data Scaling for META Ads Recommendation - MERGE: Next-Generation Item Indexing Paradigm for Large-Scale Streaming Recommendation (MERGE) - MetaPlate: Counterfactual-Guided RAG-LLM Tool for Personalized Food Recommendation and Hyperglycemia Prevention - Mind the Gap: Bridging Behavioral Silos with LLMs in Multi-Vertical Recommendations - MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music - NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems (NOVA) - Ocean4Rec: Offline LLM-Derived OCEAN Profiles for Request-Time VOD Reranking - OneBar: An End-to-End Content-Grounded Generative Query Recommendation Framework for E-Commerce Video Feeds (OneBar) - OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation (OneLive) - OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System (OnePiece) - OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework (OneSearch-V2) - PIT: A Dynamic Personalized Item Tokenizer for End-to-End Generative Recommendation (PIT) - Quantizing Intent: Cross-Domain Semantic IDs from Organic Activity for Industrial Ranking - Query-Conditioned Generative Search with QGS in Quark - Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models - Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language (Restoring Collaborative) - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - RankGraph-2: Lifecycle Co-Design for Billion-Node Graph Learning in Recommendation (RankGraph-2) - Representation Curriculum: Stagewise Training for Robust Ranking and Allocation - Scaling Dense Retrieval with LLM-Annotated Training Data: Structured Mining and Progressive Curriculum for E-Commerce Sponsored Search - Rethinking Sales Lead Scoring with LLM-based Hierarchical Preference Ranking (HPRO) - RGD: Reward Guided Decoding for Generative Recommendation (RGD) - RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems (RankUp) - SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress (SIGMA) - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - SSR: Beyond Dense Connectivity: Explicit Sparsity for Scalable Recommendation (SSR) - SIREN: Unified Multi-Granularity Semantic Interaction for Multi-Modal Lifelong User Interest Modeling - SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders (SORT) - SPARC: Sequence-aware Progressive Attribute Routing and Compression Framework for Generative Recommendation (SPARC) - Structuring and Tokenizing Distributed User Interest Context for Generative Recommendation (G2Rec) - SSRLive: Live Streaming Recommendation with Dynamic Semantic ID (SSRLive) - Synthetic Data from Cross-Domain Events for Large-Scale Recommendation Systems (SCALR) - Taiji: Pareto Optimal Policy Optimization with Semantics-IDs Trade-off for Industrial LLM-Enhanced Recommendation (Taiji) - The Case Against Generation for Retrieval: Discriminative Language Models as Effective Retrievers (Discriminative Retrieval) - Token Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models (Token Factory) - ToolRec: Calibrated Preference Alignment for Query Recommendation in On-Device Assistants (ToolRec) - Farewell to Item IDs: Unlocking the Scaling Potential of Large Ranking Models via Semantic Tokens (TRM) - Fast and Feasible: Permutation-based Constrained Reranking for Revenue Maximization (PermR) - Feedback-Grounded Policy Discovery for Generative Recommendation (Feedback-Grounded) - UniNote: A Unified Embedding Model for Multimodal Representation and Ranking - UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale - UniMixer: A Unified Architecture for Scaling Laws in Recommendation Systems (UniMixer) - Unified Multi-Task Relevance Modeling for E-Commerce: Comparing Task Routing Architectures Across LLMs and Cross-Encoders - Variance Reduction for Heavy-Tailed Monetization Metrics in Ranking Experiments via Post-Stratification (ShareChat/SIGIR 2026) - DREAM: Dynamic Refinement of Early Assignment Mappings (DREAM) - VCG: A Multimodal Retrieval Framework for E-Commerce Video Feeds under Extreme Cold-Start Conditions (VCG) - Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale (RaG) - TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems (TokenMinds) - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - From Bootstrapping to Sequence Modeling: A Unified Generative Framework for Personalized Landing-Page Modeling (GLAN) - From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations - NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems (NOVA) - IntentTune: Using User Demand and Personalization to Resolve "Unknown" Query Intents for E-commerce Search (IntentTune) - Prompt Generation Technical Report (PG) - An LLM-powered Agentic Recommendation System for Connected TV Content Discovery - Serving the Long Tail: Training-Free LLM Candidate Generation for Vacation Rental Marketplaces - Not Only NTP: Extending Training Signal Coverage for Generative Recommendation (NONTP) — Meituan, arXiv 2607.12277 - MESH: Scaling Up Retrieval with Heterogeneous Content Unification (MESH) - PinEqualizer: Full Funnel Content Exploration and Debiasing System at Pinterest (PinEqualizer) - Probabilistic Residual Learning for Online Recommendations (PRL) - RankGraph-2: Lifecycle Co-Design for Billion-Node Graph Learning in Recommendation (RankGraph-2) - GRACE: Bringing GRACE to Recommendation: Fine-Tuning for Sustainable and Accurate Personalization (GRACE) - STAR: Structured Tokenization and Target-Aware Interest Representation for PCVR Prediction (STAR) - DrEM: Dual-Side Robust Ensemble Ranking from Noisy User Preference Predictions in Video Recommendation (DrEM) - Sona Technical Report (Sona) - Improving Item Discoverability in e-Commerce Search via Related Intent Generation - Hypothesis-Driven Shelf Generation for Personalised Recommendation - Gwhere: Guess Where You Go — Generative Next Point-of-Interest Recommendation in Amap (Gwhere) - Decoupled Temporal Encoding for Generative Recommendation (DTE) - SAGA: Structure-Attended Generative Action Embedding Model that encodes Multi-Surface User Action Sequences (SAGA) - PriCoRec: A Privacy-Aware Cloud-Device Collaborative Framework for Ad Recommendation under Feature Constraints (PriCoRec) - Hierarchical Exponential-Gaussian Mixtures for Watch-Time Distribution Prediction (HEGM) - TAGR: Temporally Adaptive Generative Recommendation for Industrial Live-Streaming Advertising (TAGR) - RecGPT-Mobile-V2 Technical Report - An Event is Worth One Token: Event Tokenization for Industrial-scale LLM Recommendation (AMBER) - TransRetrieval: Scaling Up Transformer-Based Retrieval for Industrial Recommendation (TransRetrieval) - DCEO: Direct Causal Effect Optimization for Long-Term User Value Modeling in E-commerce Search (DCEO) - Stageboost: Recommending Signals Based on Counterfactual Estimation - Token-Level Advertising - HubMixer: Progressive Latent Hub Mixing for Parameter-Efficient Feature Interaction in Recommendation (HubMixer) - TGR: Tencent Generative Recommendation — Unified Generation and Reasoning (TGR) - From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs (ReST) - DMRL: Document-Mediated Reinforcement Learning for Skill Optimization in Advertising Recommendation (DMRL) - ICEGR: An Intent-Coherent End-to-End Generative Retrieval Framework for E-commerce Search (ICEGR) - LLM4AIGQ: LLM-based AI Guidance Query Generation Framework for Multi Interest Mining (LLM4AIGQ) - RecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender Systems (RecEvolve) - CAMIE: Co-Engagement-Aware Multimodal Item Embeddings for Snap Dynamic Product Ads Retrieval (CAMIE) - Personalized Task Dependency Graphs for Mitigating Signal Erosion in Multi-Task Recommendation (PTDG) - Beyond Co-purchase Relation: Evolution of Complementary Recommendations at Allegro (AlleCompanion) - SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce (SAM-D2Q) - AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems (AutoLR) - RegionFed: Federated Learning for Personalized Query Understanding in Heterogeneous Retail Environments (RegionFed) - Inventory-Grounded Policy-Level Optimization for Training-Free AI Search (IGPO) - SequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching (SequenceO1) - Agentic ML Exploration (A-MLE) for Ads Ranking (A-MLE) - UniRec: Cross-stage Multi-Task Fusion with Preference Alignment for Cascaded Recommender Systems (UniRec) - Auto-RecSys: Harnessing Autonomous Research Agents for Industry-Scale Recommender System (Auto-RecSys) - Task-Blind No MORE: Multi-Task Information Flow in Unified Ranking Backbones (MORE) - ChronicleRec: Pre-training Temporally Anchored Tokens for Lifelong User Modeling (ChronicleRec) - MIMA: Multi-Interest Recommendation via Multi-Positive Exclusive Assignment (MIMA) - LazFormer: Scaling Transformers for Industrial Recommendation via Transferable Generative Pre-training (LazFormer) - VARG: Value-Aware and Ranking-Aligned Generative Retrieval for Dynamic E-commerce Search (VARG) — Taobao & Tmall - PinDCO: Whole-Page Aware Dynamic Creative Optimization at Scale (PinDCO) — Pinterest (RecSys 2026) - PCap: Personalized Retrieval-Stage Diversity Capping in Facebook Marketplace (PCap) — Meta - Balancing Trial and Reorder: A Hybrid Sequential Transformer-GBDT Ranker for On-Demand Delivery (UVR) — Wolt / DoorDash - AURA: Agentic Diagnosis and Refinement for Production Recommender Systems at Scale (AURA) — The Walt Disney Company - LIGE-GR: A Smooth Leap from Ranking to Generative Recommendation in the LLM Era (LIGE-GR) - Scaling Articulated Rationales for MLLM-based Recommendation (SARA) - One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations (ANGLE) - Single-Token Expected-Value Scoring for Cold-Start Candidate Ranking - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 - Dense Feature Representation over Sequence Modeling: A Solution to the KDD Cup 2026 UniRec Challenge (KDD Cup 2026 UniRec) — Z Lab (Chengdu), KDD Cup 2026 UniRec Workshop - CORAL: An LLM-Native Harness for Production Recommender Systems (CORAL) — Meta AI, arXiv 2609.02730 - Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation (CRS-Performance) - The Lifecycle of LLM-as-a-Judge for Large-Scale Recommendation Explanations - From Prompting to Behavioral Alignment: Personalized LLM Judges for Recommendation Evaluation - Beyond Interleaving: Causal Attention Reformulations for Generative Recommender Systems - MuSeR: Scalable Long-sequence Recommendation with Multi-interest Modeling (MuSeR) — Baidu / CityU HK / CUHK; hierarchical temporal compression + multi-interest + beam-search retrieval, deployed at Baidu - T-RoPE: Time-Aware Rotary Position Embedding for Sequential Recommendation (T-RoPE) — Shopify, arXiv 2609.30576 - KuaFu: Compressing Long User Behavior into Understanding at Billion Scale (KuaFu) — Tencent, arXiv 2609.31045 - GRP: Snap's Generative Recommendation Paradigm (GRP) - Snap Inc., arXiv 2609.36688 - HELIX: Purified and Unified Large-Scale Recommendation (HELIX) - TikTok, arXiv 2609.37183 - Optimizing Effective Training Time for Large-Scale Recommendation Systems — Meta, arXiv 2610.02057 - When LLM-Inferred User Context Adds Value in Production Streaming Recommendation (LLM-Inferred-Context) — DePaul University / Comcast, arXiv 2609.38999 - Efficient Clustering with Quality Guardrails for LLM-based Recommender Systems at Industry Scale — Amazon, arXiv 2607.19704 - WeGR: Unified Generative Retrieval for Industrial Search Systems in WeChat (WeGR) — Tencent (WeChat), arXiv 2610.16979 - TacGR (UniSGR-V2): Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation (TacGR) — Alibaba International, arXiv 2610.16132 - TGR-EvoDAG: A DAG-based Multi-Agent Code Evolution Framework for Generative Recommendation on Tencent (TGR-EvoDAG) — Tencent / HK PolyU / McGill, arXiv 2610.32326 ### Infrastructure / Serving - FlashVector: Agent for Hierarchical Model Serving Stack Optimization (FlashVector) — Stanford University / Unity - GRACE: Generative Recommender Acceleration Engine for Real-Time Ads Retrieval (GRACE) - Enabling High-Bandwidth Flash for Generative Recommendation Serving with Write-Aware KV Cache Policy - Efficient Swing Computation for Retrieval in Large-Scale Recommender Systems (ASC / K-ASC) - Quanta: A Self-Contained Python Library for Hybrid Retrieval over Quantised Embeddings, Lexical Indexes, and Knowledge Graphs (Quanta) - SELF-INDEX: Self-Evolving Search Index (SELF-INDEX) — Yonsei University, arXiv 2609.19656 ### Inference - CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring (CompRank) - Asymmetric Generative Recommendation via Multi-Expert Projection and Multi-Faceted Hierarchical Quantization (AsymRec) - Gated Bidirectional Linear Attention for Generative Retrieval (GBLA) - Position-Aware Drafting for Inference Acceleration in LLM-Based Generative List-Wise Recommendation (PAD-Rec) - OpenOneRec - RelayGR - [STATIC] Vectorizing the Trie - MTServe: Efficient Serving for Generative Recommendation Models - Semantic Trimming and Auxiliary Multi-step Prediction for Generative Recommendation (STAMP) - One Pool, Two Caches: Adaptive HBM Partitioning for Accelerating Generative Recommender Serving - RcLLM: Accelerating Generative Recommendation via Beyond-Prefix KV Caching - Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval - Birds of a Feather Cluster Nearby: A Proximity-Aware Geo-Codebook for Local Service Recommendation - Conditional Memory Enhanced Item Representation for Generative Recommendation (ComeIR) - Recommender System as Slow and Fast Thinkers (DS-Frame) ### Inference Acceleration / Distillation - BAHSD: Bridging the Long-tail Gap via Adaptive Distillation in Black-box Sequential Recommendation (BAHSD) - Empowering Compact LLMs with Fusion of Layer-wise Exits for Recommendation (FLEXRec) - MLPs are Efficient Distilled Generative Recommenders (SID-MLP) - CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring (CompRank) - FOSTER: First-order Dataset Distillation for Text-based Sequential Recommendation - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - miniReranker: Efficient Multimodal Reranking through Visual Cache Reuse and Interaction Sparsity - MLPs are Efficient Distilled Generative Recommenders (SID-MLP) - Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models - SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders (SORT) - Empowering Cross-Domain Sequential Recommendation with Hybrid Tokenization and Serial-Parallel Decoding (GenCDSR) - DoPR: Reusable Compressed Document Prefixes for Efficient LLM Reranking (DoPR) - DuoLM: Fast and Effective Personalized Recommendation with Language Models (DuoLM) — Google DeepMind, arXiv 2610.24481 - OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models (OneLatent) — Kuaishou Technology, arXiv 2610.30675 - Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation (LoopRec) — Renmin University of China, arXiv 2610.30128 [opensource] ### Knowledge Distillation - BAHSD: Bridging the Long-tail Gap via Adaptive Distillation in Black-box Sequential Recommendation (BAHSD) - SCoRD: Semantic-Assisted Continual Retriever-Reranker Distillation for LLM-Based Recommendation (SCoRD) - SmartGR: Hierarchy and Beam-Aware Knowledge Distillation for Generative Recommendation (SmartGR) - Sona Technical Report (Sona) - Improving Item Discoverability in e-Commerce Search via Related Intent Generation - Hypothesis-Driven Shelf Generation for Personalised Recommendation - GOD: Enhancing Generalization via Deep Grafting for Sequential Recommendation (GOD) - D3ER: Supporting Multi-Modal Recommendation via Disentangle and Distillation-based Dynamic Ensemble (D3ER) - SelfDR: Self-Distillation from Reasoning for LLM-Based Recommendation (SelfDR) - Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation - P3Rec: Distilling Prior-Posterior Preference Reasoning for LLM-based Recommendation (P3Rec) - GEAR: Generative End-to-end Ad Retrieval at Douyin (GEAR) — ByteDance (Douyin); arXiv 2609.39327 - Exploring Forum Post Retrieval with Generative Modeling — Meta (Facebook Forum); arXiv 2609.38646 - KUAISHOU Explorer LLM-Rec Challenge 2026: Reasoning Generative Recommendation — Kuaishou; SIGIR 2026 challenge; arXiv 2609.39828 - OneChord: Distill What You Rank, Recommend in One Pass (OneChord) — University of Chinese Academy of Sciences, arXiv 2610.25805 - Knowledge Distillation across Item and Code Spaces for Parallel Generative Recommendation — The Hong Kong University of Science and Technology / USTC, arXiv 2610.32943 - MuSID-KD: Multigranularity Knowledge Distillation for Generative Recommendation with Semantic IDs (MuSID-KD) — East China Normal University, arXiv 2610.35218 ### LLM-as-Judge - Statistically Reliable LLM-Based Ranking Evaluation via Prediction-Powered Inference (PRECISE) - The Utility of LLMs in Recommender Systems Explanation Evaluation - Should I Be Polite to My LLM Relevance Judge? Tone as a Severity Operating-Point Shift - The Lifecycle of LLM-as-a-Judge for Large-Scale Recommendation Explanations - From Prompting to Behavioral Alignment: Personalized LLM Judges for Recommendation Evaluation - Robust Fusion of Semantic and Behavioural Signals for LLM Reranking in Personalised Search — Spotify; USRW Workshop @ RecSys 2026, arXiv 2609.25825 ### LLM-based Recommendation - RecSys Factory: Bounding LLM Agent Autonomy to Decision Points in the Industrial Recommender Lifecycle (RecSys Factory) — Tencent, arXiv 2608.11241 - ATLAS: Learning to Recommend Across Unseen Domains (ATLAS) — Sony Research India, arXiv 2608.03899 - Decision4Rec: A Jev-Inspired Framework of Decision Models on Knowledge and Active Graphs for Multi-Agent Recommender Systems (Decision4Rec) — University of Bari Aldo Moro, arXiv 2610.26035 [opensource] - Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems — Pinterest, arXiv 2609.33073 - Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops — Spotify, arXiv 2609.30297 - Do Evidence-Reading Diagnostics Improve Interface Selection in Small LLM Recommenders? — Independent Researcher, arXiv 2609.37472 - Overview and Analysis of the RecSys Challenge 2026: Conversational Music Recommendation — Sony Group Corporation, arXiv 2609.33045 - The Recall Ceiling of LLM Recommendation Reranking (RecallCeiling) — University of Southern California, arXiv 2609.27953 - LSF-SR: Latent Semantic Fusion for Sequential Recommendation via Flow-based Conditional Variational Autoencoders (LSF-SR) — National Yang Ming Chiao Tung University, arXiv 2609.29815 - When LLM-Based User Profiling Adds Value in Production Streaming Recommendation (WhenLLM-Profiling) — DePaul University - Who You Are Adds Nothing Detectable to Where You Go Next: Sociodemographic Conditioning in LLM Next-Location Prediction (Next-Location) — University of Melbourne - CogRec: Structure-Cognitive Fast-and-Slow Reasoning for Generative Recommendation (CogRec) - Discriminative Language Models as Effective Retrievers (Discriminative Retrieval / Case Against GenRec) - EGR: Embedding-Native Generative Retrieval with a Shared LLM (EGR) - Empowering Compact LLMs with Fusion of Layer-wise Exits for Recommendation (FLEXRec) - GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling (GateDiffInt) - Grevo: A Unified Generative Recommendation Framework with Evolutionary Item Indexing (Grevo) - LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation (LaRec) - OxygenREC-v2: Internalizing Discrimination into Generative Recommendation (OxygenREC-v2) - POI Recommendation with LLM-Augmented Multi-Graph Learning and Contrastive Alignment (LLM-MGCL) - Profiling What Matters: Context-Aware Item Profiles from Large-Scale Metadata for LLM Recommenders (CAIRO) - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - rEDMRec: Distilling Large Language Model Reasoning into an Editable Experience Memory for Recommendation (rEDMRec) - SCoRD: Semantic-Assisted Continual Retriever-Reranker Distillation for LLM-Based Recommendation (SCoRD) - TopoGR: Revealing and Preserving Latent Structure of Semantic ID in Generative Recommendation (TopoGR) - Training-Free LLM-Based Recommendation with Post-LLM Item Refinement Using Collaborative Signals (CoRRe) - VaLiDRec: Variable-Length LLM-Aligned Semantic IDs for Generative Recommendation (VaLiDRec) - ACE: Anisotropy-Controllable Embedding for LLM-enhanced Sequential Recommendation - APAO: Bridging the Training-Inference Gap in Generative Recommendation via Adaptive Prefix-Aware Optimization (APAO) - APG4RecSim: Task-Aware Automated User Profile Generation - Atomic Intent Reasoning: Bringing LLM Semantics to Industrial Cross-Domain Recommendations (AIR) - Beyond Matching: Category-Guided Latent Intent Reasoning for Generative Retrieval in E-Commerce (CaLIR) - BRIDGE: Behavior-Guided Candidate Calibration for Multimodal Recommendation - Bridging Short Videos and Live Streams: Reasoning-Guided Multimodal LLMs for Cross-Domain Representation Learning (RGCD-Rep) - Can LLM Rerankers Predict Their Own Ranking Performance? - Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation (CausalDPO) - Causal-Invariant Cross-Domain Out-of-Distribution Recommendation (CICDOR) - Causal Representation Learning for Generalisable Recommendation - CFALR: Collaborative Filtering-Augmented Large Language Model for Personalized Fashion Outfit Recommendation (CFALR) - Cold-Start Personalization via Training-Free Priors from Structured World Models (PEP) - CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring (CompRank) - Consensus vs. Dissent: Dynamic LLM Modeling of Subjective Preferences in Group Recommenders - DeGRe: Dense-supervised Generative Reranking for Recommendation - Designing Trustworthy LLM-based Wellbeing Recommendation through Controllable Interaction - Dialogue to Discovery: Attribute-Aware Preference Elicitation for Conversational Product Search Assistants (D2D) - Diffusion-GR2: Diffusion Generative Reasoning Re-ranker (Diffusion-GR2) - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - Dual-Stream MLP is All You Need for CTR Prediction (DS-MLP) - Every Preference Has Its Strength: Injecting Ordinal Semantics into LLM-Based Recommenders (OSA) - EviRank: Evidence-Based Confidence Estimation for LLM-Based Ranking (EviRank) - Federated Variational Preference Alignment with Gumbel-Softmax Prior for Personalized User Preferences (FedVPA-GP) - Federated User Behavior Modeling for Privacy-Preserving LLM Recommendation (SF-UBM) - Feedback-Grounded Policy Discovery for Generative Recommendation (Feedback-Grounded) - LLM-HYPER: Generative CTR Modeling for Cold-Start Ad Personalization via LLM-Based Hypernetworks (LLM-HYPER) - Let the Agent Steer: Closed-Loop Ranking Optimization via Influence Exchange (Sortify) - Self-Distilled Reinforcement Learning for Co-Evolving Agentic Recommender Systems (CoARS) - Filling the Gaps: Selective Knowledge Augmentation for LLM Recommenders (KnowSA_CKP) - LLM-Based Re-Ranking for Real Estate Search - Think When Needed: Model-Aware Reasoning Routing for LLM-based Ranking - From Hidden Profiles to Governable Personalization: Recommender Systems in the Age of LLM Agents - From Prompt to Purchase: How AI Brand Recommendations Move Consumers on the Open Web - From "Strings" to "Things" for Personal Knowledge Graphs (PKG-Rec) - GARDRec: Decision-Level Graph Grounding for LLM Recommendation (GARDRec) - GCRS: Generative Conversational Recommender System - GEMs: Breaking the Long-Sequence Barrier in Generative Recommendation with a Multi-Stream Decoder (GEMs) - Gender and Race Bias in Consumer Product Recommendations by Large Language Models - Generative Archetype-Grounded Item Representations for Sequential Recommendation (GenAIR) - Generative Reasoning Re-Ranker (GR2) - GenPage: Towards End-to-End Generative Homepage Construction at Netflix (GenPage) - GLIDE: Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify (GLIDE) - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - GraphLoRA: Structure-Aware Low-Rank Adaptation for Large Language Model Recommendation (GraphLoRA) - Grevo: A Unified Generative Recommendation Framework with Evolutionary Item Indexing (Grevo) - Grounded Token Initialization for New Vocabulary in LMs for Generative Recommendation (GTI) - Guiding Federated Graph Recommendation with LLM-encoded knowledge - Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation (ReaEmb) - HiLaR: Hierarchical Latent Reasoning for LLM-based Recommendation (HiLaR) - HoloRec: Holistic Encoding and Interleaved Reasoning for Generative Recommendation (HoloRec) - Implicit Reasoning for Large Language Model-based Generative Recommendation (PauseRec) - INSPIRE: Intent-aware Neural Sponsored Product Retrieval for E-commerce - Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems - Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems - SIREN (Luring LLMs onto the Rocks): PAIR-Driven Preference Manipulation in Web-RAG Recommenders (SIREN) - Intuition-Guided Latent Reasoning for LLM-Based Recommendation (IntuRec) - Joint Optimization of Relevance and Engagement in Multi-Task Ranking for E-Commerce with Efficient LLM Supervision - L2Rec: Towards Dual-View Understanding of LLMs for Personalized Recommendation - Learning User-Aware Recall: Personalized Retrieval in Long-Term Conversational Memory (PPRO) - LLM Retrieval for Stable and Predictable Ad Recommendations - An LLM-Powered Semantic Alignment Framework for Journal Recommendation - LLM-as-a-Judge for Reliable and Explainable Offline Evaluation in Top-K Recommendation (LLM Judge) - LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems - LLM-Based User Personas for Recommendations at Scale - LLMs Need Encoders for Semantic IDs Too (PrefixMem) - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - LRanker: LLM Ranker for Massive Candidates - Memento: Personalized RAG-Style Long-Retention Data Scaling for META Ads Recommendation - MetaPlate: Counterfactual-Guided RAG-LLM Tool for Personalized Food Recommendation and Hyperglycemia Prevention - Mind the Gap: Bridging Behavioral Silos with LLMs in Multi-Vertical Recommendations - MixRAGRec: MoE KG-RAG for Multi-Agent LLM Recommendation - MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music - Macro Graph of Experts for Billion-Scale Multi-Task Recommendation (MGOE) - Mult-DPO: Multinomial Direct Preference Optimization for Recommender Systems - Multimodal Music Recommendation System using LLMs - Music Playlist Captioning at Scale with Large Language Models - Ocean4Rec: Offline LLM-Derived OCEAN Profiles for Request-Time VOD Reranking - OMEGA: Collaborative Memory Augmentation for Generative Recommendation (OMEGA) - On the Memorization Behavior of LLMs in Generative Recommendation: Observations, Implications, and Training Strategies (IIRG) - One Polluted Page Is Enough: Evaluating Web Content Pollution in Generative Recommenders (FORGE) - OneBar: An End-to-End Content-Grounded Generative Query Recommendation Framework for E-Commerce Video Feeds (OneBar) - OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System (OnePiece) - PaperFlow: Profiling, Recommending, and Adapting Across Daily Paper Streams (PaperFlow) - Prompt Optimization for User Simulation in Conversational Recommender Systems (UserSimulator) - Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation (Scaling Laws) - Query-Conditioned Generative Search with QGS in Quark - RAGEAR: Retrieval-Augmented Graph-Enhanced Academic Recommender - RankGR: Rank-Enhanced Generative Retrieval with Listwise Direct Preference Optimization in Recommendation (RankGR) - Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval - RSIR: Can Recommender Systems Teach Themselves? A Recursive Self-Improving Framework with Fidelity Control (RSIR) - ReasonRec: A Reasoning-Augmented Multimodal Agent for Unified Recommendation (ReasonRec) - RECAP: Feedback-Driven Streaming Semantic User Profiles for Short-Video Recommendation (RECAP) - Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models - RecGPT-V3 Technical Report (RecGPT-V3) - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale (RaG) - ReasonRec: A Reasoning-Augmented Multimodal Agent for Unified Recommendation (ReasonRec) - Reproducing FACTER: Fairness via Conformal Thresholding and Prompt Repair - Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language (Restoring Collaborative) - Rethinking Sales Lead Scoring with LLM-based Hierarchical Preference Ranking (HPRO) - Rethinking Fairness in LLM-Based Recommender Systems: A Survey - RouteRec: Strict Evaluation of Recommender-Agent Selection and Aggregation (RouteRec) - RPORec: Reinforced Preference Optimization for Reasoning-Augmented Recommendations - SAERec: Constructing Fine-grained Interpretable Intents Priors via Sparse Autoencoders for Recommendation (SAERec) - Scaling Dense Retrieval with LLM-Annotated Training Data: Structured Mining and Progressive Curriculum for E-Commerce Sponsored Search - SAILRec: Steering LLM Attention to Dual-Side Semantically Aligned Collaborative Embeddings for Recommendation (SAILRec) - SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress (SIGMA) - Mitigating Reward Hacking in LLM-based Recommendation: A Preference Optimization Approach (SIRIUS) - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - SimGR: Escaping the Pitfalls of Generative Decoding in LLM-based Recommendation (SimGR) - Taiji: Pareto Optimal Policy Optimization with Semantics-IDs Trade-off for Industrial LLM-Enhanced Recommendation (Taiji) - TAP-PER: Beyond Retrieval: Learning Compact User Representations for Scalable LLM Personalization (TAP-PER) - TCA4Rec: Token-level Collaborative Alignment for LLM-based Generative Recommendation - The Case Against Generation for Retrieval: Discriminative Language Models as Effective Retrievers (Discriminative Retrieval) - The Injection Paradox: Brand-Level Suppression in Safety-Trained LLM Recommendations via RAG Context Injection (Injection Paradox) - ThinkGR: Integrating Chain-of-Thought into Generative Retrieval - TopoGR: Revealing and Preserving Latent Structure of Semantic ID in Generative Recommendation (TopoGR) - TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems (TokenMinds) - ToolRec: Calibrated Preference Alignment for Query Recommendation in On-Device Assistants (ToolRec) - Toward User Preference Alignment in LLM Recommendation via Explicit Context Feedback - Towards Fast Domain Adaptation and Fine-Grained User Simulation for Evaluating Conversational Recommender Systems (AdaptSim) - TRACE: A Conversational Framework for Sustainable Tourism Recommendation with Agentic Counterfactual Explanations - Trading Engagement for Sustainability: Carbon-Aware Re-ranking for E-commerce Recommendations - Trustworthy Recommendation in the Era of Large Language Models: Opportunities and Challenges - Uncertainty and Fairness Awareness in LLM-Based Recommendation Systems - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - Unified Multi-Task Relevance Modeling for E-Commerce: Comparing Task Routing Architectures Across LLMs and Cross-Encoders - UniNote: A Unified Embedding Model for Multimodal Representation and Ranking - UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale - Unifying Search and Recommendation in LLMs via Gradient Multi-Subspace Tuning (GEMS) - VaLiDRec: Variable-Length LLM-Aligned Semantic IDs for Generative Recommendation (VaLiDRec) - VirtualMLE: A Virtual ML Engineer that Optimizes Sequential Recommenders (VirtualMLE) - Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models (LLM Brand) - VENOMREC: Cross-Modal Interactive Poisoning for Targeted Promotion in Multimodal LLM Recommender Systems (VENOMREC) - Whole-Pool Setwise Reranking with Long-Context Language Models (WP-Setwise / DualEnd) - Whose Name Comes Up? III: Persona Prompting Effects in LLM-Based Scholar Recommendation - From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations - NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems (NOVA) - SocialPersona: Benchmarking Personalized Profiling and Response with Multimodal Social-Media Context (SocialPersona) - Bi-NAS: Towards Effective and Personalized Explanation for Recommender Systems via Bi-Level Neural Architecture Search (Bi-NAS) - CoPersona: Collaborative Persona Graphs for Robust LLM Personalization (CoPersona) - ExPerT: Personalizing LLM Responses to Users' Domain Expertise via Query-Wise Semantic and Keystroke Behavioral Cues (ExPerT) - Tokenizing Numerical and Embedding Features for LLM RecSys - Making Collaborative Signals Count: Graph-Aware Large Language Models for Sequential Recommendation (GALLM) - TRACER: Balancing Stability-Plasticity-Cognitivity Trilemma for LLM Enhanced Continual Recommendation (TRACER) - Ask to Be Sure: Informative Interactions for Confident Multi-Turn LLM Recommendation (Ask to Be Sure) - Rethinking Semantic Alignment in LLM-Enhanced Collaborative Filtering: A Spectral Decoupling Approach (UniSpecRec) - An Event is Worth One Token: Event Tokenization for Industrial-scale LLM Recommendation (AMBER) - MOTIF: Motivation-guided Topology Inference for Cold-start Multimodal Recommendation (MOTIF) - Beyond Ranking Accuracy: Evaluating LLM-Cited Feature Rationales for Next Basket Repurchase Recommendation - RPCBench: A Benchmark for Proactive Premise Critique in LLM-based Recommendation (RPCBench) - World Model-Guided Reinforcement Learning via Counterfactual User Engagement Simulation (WMG-RL) - DREAMS: Dual-node Monte Carlo Tree Search for Conversational Recommendation (DREAMS) - SPAR: Enhancing Industrial-Scale Generative POI Recommendation via Real-World Spatial Perception (SPAR) - LLM4AIGQ: LLM-based AI Guidance Query Generation Framework for Multi Interest Mining (LLM4AIGQ) - Latent-Aligned Reasoning for Multimodal Recommendation (LARK) - AtomRec: Evolving Atomic Memory for Agentic Recommendation (AtomRec) - Repeated Queries Exhaust an LLM's Brand Recommendations but Not Its Sources - Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation - AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems (AutoLR) - HyperTrace: Hypothesis-Based Preference Tracing for Online LLM Personalization (HyperTrace) - Agentic Share-of-Search: A Multi-Agent AI System for Competitive Decision-Making in LLM-Mediated E-Commerce (ASoS) - Generative Marketing Mix Modeling: A Causal Inference Framework Linking GEO and GEM to Business Impact (GMMM) - Generate to Explore, Select to Exploit: Aligning LLM-based Headline Generation with Personalized Recommendation (GESE) - P3Rec: Distilling Prior-Posterior Preference Reasoning for LLM-based Recommendation (P3Rec) - Safety as a Constraint: Fine-Tuning a LLM Recommender to Explain Itself - TATK: Triple-Aware Top-K Learning with Knowledge-Grounded Verification for LLM-based Sequential Recommendation (TATK) - ReliGRec: Reliability-Oriented LLM-Based Generative Recommendation via User-Risk-Aware Prompt Routing (ReliGRec) - Evaluating Brand Retrieval and Ranking in Large Language Model Recommendations - Scaling Articulated Rationales for MLLM-based Recommendation (SARA) - Single-Token Expected-Value Scoring for Cold-Start Candidate Ranking - Understanding AI Provider Recommendations in Local Service Markets - Reproducing Transparent and Scrutable Recommendations: Exploring Open-Weight Models via Natural-Language User Profiles (Transparent UPR Repro) — University of Zurich, BlackBoxNLP @ EMNLP 2026 - Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment (Re2A) — HK PolyU / CUHK / Sichuan University, EMNLP 2026 - FacetCRS: Multi-Faceted Preference Learning for Pricking Filter Bubbles in Conversational Recommender System (FacetCRS) — Sun Yat-sen University, AAAI 2024 - Epistemic Warrant for LLM Recommendations: Characterizing the Basis for Reliance When Ground Truth Is Unavailable — Purdue University / University of Pennsylvania, arXiv 2609.04127 - Explainable Recommendations at Scale: LLM Rationales for YouTube Music Artist Discovery — Google; async LLM rationale generation for exploration - REPREC: Representation Driven Parameter-Efficient Recommendation System (REPREC) — Ohio State University / Capital One; arXiv 2607.24845 - X-KGRank: A Knowledge Graph RAG Framework for Explainable Recommendations via Pattern Mining and LLM Re-Ranking (X-KGRank) — San Jose State University; arXiv 2608.01732 - RecToolBench: Benchmarking Recommendation-Specific Tool Orchestration under Fuzzy User Intent (RecToolBench) — University of Virginia / Jilin University / Squirrel AI / PolyU, arXiv 2609.30717 - KuaFu: Compressing Long User Behavior into Understanding at Billion Scale (KuaFu) — Tencent, arXiv 2609.31045 - AgentRecommender: LLM Agents Enable Customizable Recommender Systems on the User Side (AgentRecommender) — National Institute of Informatics, arXiv 2609.31166 - EvoSkillRec: LLM-Driven Skill-Genome Evolution for Recommender Architecture Discovery (EvoSkillRec) - Huawei Noah's Ark Lab / City University of Hong Kong, arXiv 2609.34552 - Not All Is Lost: Repairing Lossy User Preference States of Personalization Encoders (REPAIR) — IIIT Delhi, arXiv 2610.01270 - When More Data Is Not Enough: The Context-Sufficiency Frontier in Generative AI Personalization (Context-Sufficiency) — ABYAT, arXiv 2610.00654 - When LLM-Inferred User Context Adds Value in Production Streaming Recommendation (LLM-Inferred-Context) — DePaul University / Comcast, arXiv 2609.38999 - Decision-Oriented Recommendation Reranking: An Empirical Study of Jev (Jev) — University of Rochester / Meta AI, arXiv 2609.40241 - When History Misleads: Asymmetric Margin Supervision for Instruction-Guided LLM Generative Recommendation (AIMS) — Duke University, arXiv 2610.02600 - Reasoning with Evidence, Not Merely Rationales: Verifiable Preference Proofs for LLM-Based Recommendation (PROVE-REC) — Yonsei University, arXiv 2610.02968 - A Behavioral Trait Leaks into Preferences: Diagnosing Trait Interference in LLM User Simulators (PQA) — KAIST, arXiv 2609.25572 [opensource] - System Attribution in LLM Brand Recommendations (Dmitrij Żatuchin) — EUAS / Rankfor.AI - Learning Robust Personalized Prompts for LLM-Driven Sequential Recommendation (LRPRec) — Zhejiang University, arXiv 2610.03923 - MATE: Adaptive Long- and Short-Term User Memory for LLM-Based Recommendation (MATE) — Yonsei University, arXiv 2610.06050 - Beyond Sequences: Distilling Structured Decision Memory for LLM Recommendation (MARI) — Alibaba Group, arXiv 2610.11501 - Beyond Successor Accuracy: State Retention for Recursive Self-Improvement in Recommendation (RecRSI) — University of Hong Kong, arXiv 2610.07105 - CARA: Cognitive Adaptive Recommendation Agent — The Chinese University of Hong Kong / Tongji University, arXiv 2608.16919 - CALMRec: Causally Aligned Language Memory for Long-Horizon Recommendation — Shenzhen University, arXiv 2607.23647 - Textual User Taste: Natural-Language Taste Profiles at Spotify Scale (Textual User Taste) - Spotify, arXiv 2609.35285 - Personalized LLM Recommendation with Semantic IDs: Autoregressive Beam Search vs. Multi-Token Prediction (Netflix SID Study) — Netflix, NeurIPS 2026 PALM Workshop, arXiv 2610.15864 - SharpRec: Sharpness-aware Model Merging with Salience Recovery for LLM-based Cross-Domain Sequential Recommendation (SharpRec) — Zhejiang University / Hangzhou Dianzi University, KDD 2026, arXiv 2607.25366 - RES-MR: Risk-Aware Reasoning for Explainable and Safe Medication Recommendation (RES-MR) — Qilu University of Technology / UTS, SIGIR 2026, arXiv 2610.14474 - Vela: User Behavior as Supervision Sparks LLM Intelligence (Vela) — AllSpark at Xiaohongshu (RedNote), arXiv 2610.14709 [opensource] - QD-Rec: Open-Ended Evolution of Recommendations via Quality-Diversity (QD-Rec) — Sakana AI, arXiv 2610.23470 - LSNR: Learned Sparse News Recommendations with Attentive User Modeling (LSNR) — University of Amsterdam, SIGIR-AP 2026 - When Semantics Fall Short: Persona-Guided Cross-Domain Sequential Recommendation with Large Language Models (PerCDR) — Beijing Jiaotong University, arXiv 2610.25221 - DuoLM: Fast and Effective Personalized Recommendation with Language Models (DuoLM) — Google DeepMind, arXiv 2610.24481 - OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models (OneLatent) — Kuaishou Technology, arXiv 2610.30675 - BGAFlow: Bezier-Guided Adaptive Flow for Cross-domain Sequential Recommendation (BGAFlow) — Anhui University, arXiv 2610.30825 ### Memory Augmentation - HEM-Med: A Hierarchical Agentic Memory Framework for Safe Multi-Center Medication Recommendation (HEM-Med) — Qilu University of Technology / UTS, arXiv 2610.20191 - OMEGA: Collaborative Memory Augmentation for Generative Recommendation (OMEGA) - rEDMRec: Distilling Large Language Model Reasoning into an Editable Experience Memory for Recommendation (rEDMRec) - When Memory Takes Gradients: Collaborative Vector Memory for Agentic Recommender Systems (CoVeMem) - AtomRec: Evolving Atomic Memory for Agentic Recommendation (AtomRec) - Continual Graph Memory for Adaptive Recommendation under Intent Drift (CGM-Rec) - PRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong Conversations (PRAGMA) - Kernel-Managed Shared Memory for System-Wide Personalization - MemRetriever: Learning to Search, Reflect, and Retrieve from Long-Term Memory (MemRetriever) - Self-Evolving Memory for Generative Recommendation (LION) - LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory, with ICE v2 as an Audited Local-First Architecture (LSREP) / ICE v2 - SELF-INDEX: Self-Evolving Search Index (SELF-INDEX) — Yonsei University, arXiv 2609.19656 - Enhancing Group Recommendation with Memory-Augmented Reasoning in LLM Agent (AGR) - IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation MoE (IntBMoE) — Alibaba (AMap); deployed in AMap generative rec, arXiv 2609.21346 - ReMem: Time-Evolving Dynamic Memory for Recommendation Agents (ReMem) - Hong Kong Polytechnic University / NTU, arXiv 2609.37311 - TIDE: How Can Recommendation Feedback Evolve Agent Memory? (TIDE) - Harbin Institute of Technology / Alibaba / Institute of Automation CAS, arXiv 2609.37544 ### MoE - Generative Large-Scale Pre-trained Models for Automated Ad Bidding Optimization - L2Rec: Towards Dual-View Understanding of LLMs for Personalized Recommendation - MixRAGRec: MoE KG-RAG for Multi-Agent LLM Recommendation - OneMall - RouteRec: Behavior-Guided Sparse Routing for Sequential Recommendation (RouteRec) — KAIST / Seoul National University, arXiv 2609.39007 [opensource] ### Model / Architecture - Affective Music Recommendation: A Rollout-Based World Model for Offline Preference Optimization (AMRS) - Beyond Instance-Level Alignment and Uniformity: Semantic Factor Learning for Collaborative Filtering (SaFeAU) - Causal Representation Learning for Generalisable Recommendation - DeepInterestGR: Mining Deep Multi-Interest Using Multi-Modal LLMs - Dual-Stream MLP is All You Need for CTR Prediction (DS-MLP) - Empowering Compact LLMs with Fusion of Layer-wise Exits for Recommendation (FLEXRec) - Generative Recommendation for Large-Scale Advertising (GR4AD) - GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation - IntRR: A Framework for Integrating SID Redistribution and Length Reduction - L2Rec: Towards Dual-View Understanding of LLMs for Personalized Recommendation - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - LRanker: LLM Ranker for Massive Candidates - LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction (LoopCTR) - MiniOneRec - MixRAGRec: MoE KG-RAG for Multi-Agent LLM Recommendation - Multi-Business Prediction for Generative Recommendation at Meituan (MBGR) - One Model, Two Markets: Bid-Aware Generative Recommendation (GEM-Rec) - One Pool, Two Caches: Adaptive HBM Partitioning for Accelerating Generative Recommender Serving - OneFeed: A Unified Generative Framework for Feed Content Enhancement and Query Generation (OneFeed) - OneMall - OpenOneRec - Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models - RecGPT-Mobile: On-Device Large Language Models - SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders (SORT) - TAP-PER: Beyond Retrieval: Learning Compact User Representations for Scalable LLM Personalization (TAP-PER) - TCA4Rec: Token-level Collaborative Alignment for LLM-based Generative Recommendation - Token Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models (Token Factory) - UniDot: A Unified Network for Sequence Modeling and Feature Interaction in Large-scale Recommendation (UniDot) - UniGRec - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - UniRec: Bridging the Expressive Gap via Chain-of-Attribute - UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence - NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems (NOVA) - An Event is Worth One Token: Event Tokenization for Industrial-scale LLM Recommendation (AMBER) - Hamiltonian Spectral-Temporal Dissipative Dynamics for Sequential Recommendation (HSR) - Beyond a Single Story: Meta-Reviewing Sparse and Incomplete User-generated Contents for Recommendation (MOSAIC) - HubMixer: Progressive Latent Hub Mixing for Parameter-Efficient Feature Interaction in Recommendation (HubMixer) - Information-Guided Selective Modality-Interest Alignment for Multimodal Recommendation (AMUR) - From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs (ReST) - Beyond Modality Harmony: Orthogonal Purification and Topology-Guided MoE for Conflict-Aware Multimodal Recommendation (OrthoRec) - From Feature Interaction to Feature Transport - A Unified Block for Scalable Recommendation Models (CRAFT) - On the Regularization Landscape for the Linear Recommendation Models - UniRec: Cross-stage Multi-Task Fusion with Preference Alignment for Cascaded Recommender Systems (UniRec) - Beyond Interleaving: Causal Attention Reformulations for Generative Recommender Systems - HELIX: Jointly Scaling Feature Interaction and Sequence Modeling (HELIX) - TikTok, arXiv 2609.37183 - EvoSkillRec: Skill-Genome Evolution for Recommender Architecture Discovery (EvoSkillRec) - Huawei Noah's Ark Lab / City University of Hong Kong, arXiv 2609.34552 - RecSoup: Efficient Model Merging Approach for Controllable Multi-Objective Recommendation (RecSoup) — Renmin University of China, CIKM 2026 ### Multi-behavior Recommendation - BITRec - Dynamic Spectral Denoising with Global-Context Attention for Multi-Behavior Recommendation (SpectraMB) - GCIB: Graph Contrastive Information Bottleneck for Multi-Behavior Recommendation - PHKT: Personalized Dynamic Hypergraph-enhanced KAN-Transformer for Multi-behavior Sequential Recommendation (PHKT) - Beyond Observed Auxiliary Relations: Environment-Conditioned Modeling for Multi-Behavior Recommendation (BOAR) - MIMA: Multi-Interest Recommendation via Multi-Positive Exclusive Assignment (MIMA) - FatsMB: From Agnostic to Specific: Latent Preference Diffusion for Multi-Behavior Sequential Recommendation (FatsMB) — CAS / Kuaishou, arXiv 2602.23132 ### Multimodal - Anatomy of a Decision: Uncertainty-aware Hierarchical Intent Learning via Flow Matching for Multimodal Recommendation (UHIFlow) — Northeastern University, China, arXiv 2609.29609 - MM-VeriRec: Failure-Guided Fusion for Verifiable Agentic Multimodal Recommendation — Independent Researcher, arXiv 2609.31718 - NarraLite: Efficient Multimodal Generative Recommendation with Latent Narrative Reasoning (NarraLite) — Tencent / HK PolyU - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - UnpairGR: Unpaired Modality-Agnostic Generative Recommendation (UnpairGR) - A Cascaded Generative Approach for e-Commerce Recommendations - A General Framework for Multimodal LLM-Based Multimedia Understanding in Large-Scale Recommendation Systems (Multimodal LLM Rec) - A2Gen (Action-Aware Generative Sequence Modeling) - Agentic Recommender System with Hierarchical Belief-State Memory (MARS) - Beyond Patches: Superpixel Token-based Transformers for Attribute-Specific Fashion Retrieval (SuperFashion) - Beyond Unimodal Boundaries: Generative Recommendation with Multimodal Semantics (MGR-LF++) - Beyond Static Best-of-N: Bayesian List-wise Alignment (BLADE) - BRIDGE: Behavior-Guided Candidate Calibration for Multimodal Recommendation- Modality-Aware Identity Construction and Counterfactual Structure Learning for ID-Free Multimodal Recommendation (MAIL) - Bridging Behavior and Semantics for Time-aware Cross-Domain Sequential Recommendation (BST-CDSR) - Bridging Short Videos and Live Streams: Reasoning-Guided Multimodal LLMs for Cross-Domain Representation Learning (RGCD-Rep) - Bringing Reasoning to Generative Recommendation Through the Lens of Cascaded Ranking (CARE) - CFALR: Collaborative Filtering-Augmented Large Language Model for Personalized Fashion Outfit Recommendation (CFALR) - Closing the Indexing-Decoding Gap in Multimodal Generative Retrieval via Prefix Retention Optimization (PRO) - Conditional Memory Enhanced Item Representation for Generative Recommendation (ComeIR) - DC4SR (Disagreement as Signals) - DeepInterestGR: Mining Deep Multi-Interest Using Multi-Modal LLMs - Discrimination Is Generation (DIG) - Dual-Diffusional Generative Fashion Recommendation (DualFashion) - DUET: Joint Exploration of User Item Profiles in Recommendation System - Evaluating Scene-based In-Situ Item Labeling for Immersive Conversational Recommendation (ICRS) - Enhancing Local Life Service Recommendation with Agentic Reasoning in Large Language Model - F-GRPO: Factorized Group-Relative Policy Optimization - Factorized Latent Reasoning for LLM-based Recommendation (FLR) - FusID: Modality-Fused Semantic IDs for Generative Music Recommendation (FusID) - Generative Sequential Recommendation via Hierarchical Behavior Modeling (GAMER) - GRLM (Structured Term IDs) - HSUGA: LLM-Enhanced Recommendation with Hierarchical Semantic Understanding - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation - LASAR: Latent Adaptive Semantic Aligned Reasoning for Generative Recommendation - MARC (Modular Representation Compression) - miniReranker: Efficient Multimodal Reranking through Visual Cache Reuse and Interaction Sparsity - MLTFR (Multi-LLM Token Filtering) - Multimodal Music Recommendation System using LLMs - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - ORBIT: Preserving Foundational Language Capabilities in GenRetrieval via Origin-Regulated Merging - Popcorn: A Configurable Benchmark for Visual Evidence in Multimodal Movie Recommendation (Popcorn) - PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation - ProMax: Exploring the Potential of LLM-derived Profiles - RecGPT-Mobile: On-Device Large Language Models - RecRM-Bench: Benchmarking Multidimensional Reward Modeling - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale (RaG) - ReRec - Rethinking Semantic Collaborative Integration - RRCM: Ranking-Driven Retrieval over Collaborative and Meta Memories for LLM Recommendation - SIREN: Unified Multi-Granularity Semantic Interaction for Multi-Modal Lifelong User Interest Modeling - SSRLive: Live Streaming Recommendation with Dynamic Semantic ID (SSRLive) - Teach Multimodal Recommendation Model to See via Personalized Visual Extraction and Adaptive Learning (REVEAL) - TriAlignGR: Triangular Multitask Alignment with Multimodal Deep Interest Mining for Generative Recommendation - TwiSTAR: Think Fast, Think Slow, Then Act, Generative Recommendation with Adaptive Reasoning - UniNote: A Unified Embedding Model for Multimodal Representation and Ranking - URecJPQ: Memory-efficient Multimodal Recommendation Models through RecJPQ in Large-Scale Scenarios (URecJPQ) - SocialPersona: Benchmarking Personalized Profiling and Response with Multimodal Social-Media Context (SocialPersona) - VCG: A Multimodal Retrieval Framework for E-Commerce Video Feeds under Extreme Cold-Start Conditions (VCG) - Stream-aware Side Adaptation for Large Pre-trained Multimodal Embedding Models in Sequential Recommendation (Stresa) - VRec - When Text-as-Vision Meets Semantic IDs in Generative Recommendation: An Empirical Study - Why Users Go There: World Knowledge-Augmented Generative Next POI Recommendation - PVTG: Personalized Video Thumbnail Generation with Preference-aware Highlight Retrieval (PVTG) - Generative Universal Multimodal Retrieval with Dual-role Identifiers (DrIG) - TimeRoute: Time-Aware Modality Routing and Diffusion for Multi-Modal Recommendation (TimeRoute) - Gwhere: Guess Where You Go — Generative Next Point-of-Interest Recommendation in Amap (Gwhere) - D3ER: Supporting Multi-Modal Recommendation via Disentangle and Distillation-based Dynamic Ensemble (D3ER) - MOTIF: Motivation-guided Topology Inference for Cold-start Multimodal Recommendation (MOTIF) - SG-UMP: Sequence-Guided Universal Multimodal Prioritization Calculation Framework (SG-UMP) - Information-Guided Selective Modality-Interest Alignment for Multimodal Recommendation (AMUR) - Beyond Modality Harmony: Orthogonal Purification and Topology-Guided MoE for Conflict-Aware Multimodal Recommendation (OrthoRec) - Agents as Knowledge Integrator and Utilizer in Multimodal Recommendation (AgentMMRec) - CAMIE: Co-Engagement-Aware Multimodal Item Embeddings for Snap Dynamic Product Ads Retrieval (CAMIE) - Latent-Aligned Reasoning for Multimodal Recommendation (LARK) - MURAL: Multimodal Uncertainty-aware Recommendation via Adaptive edge Learning (MURAL) - SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce (SAM-D2Q) - SAGE: Semantic Attribute Graphs for Multi-Entity Visual Retrieval (SAGE) - RegRet: Enhancing Region-Level Retrieval in Large Multimodal Models (RegRet) — ECCV 2026 - Scaling Articulated Rationales for MLLM-based Recommendation (SARA) - Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment (Re2A) — HK PolyU / CUHK / Sichuan University, EMNLP 2026 - AdaM-Rec: Adaptive Modality Routing for Multimodal Recommendation (AdaM-Rec) — University of Queensland, arXiv 2609.38455 [opensource] - Attribute-Conditioned Multimodal Slot Factorization for Controllable Fashion Retrieval (MM-slotgate) — Amazon, arXiv 2608.12570 - Reading the Mood: Emotion-Guided Book-to-Music Recommendation via CGANs and LLMs (SAGA-CDR) — National Technical University of Athens, SENTIRE 2026 (ICDM Workshops), arXiv 2610.06703 - JBM-Diff: Joint Behavior-guided and Modality-coherence Conditional Graph Diffusion Denoising for Multi-Modal Recommendation (JBM-Diff) — Huazhong University of Science and Technology, arXiv 2604.03654 - Multimodal Graph Retrieval-Augmented Sequential Recommendation via Collaborative Filtering Paths (MGRASRec) — University of New South Wales, arXiv 2610.11228 - Seeing the Context: Image-Derived Contextual Signals for Recommendation (Seeing the Context) - Tel Aviv University, arXiv 2610.08407 - OpticalRec: Treating Text as Visual Glyphs for Multimodal Recommendation (OpticalRec) - UC San Diego, arXiv 2610.05432 - Autoregressive Retriever: Improving Query Understanding from Item Feedback for Universal Multimodal Retrieval (ARR) -- Beijing Institute of Technology / Xiaohongshu; arXiv 2610.11666 (autoregressive retrieval loop with item-feedback refinement via SFT + RL) - MLLMRec-R1: Incentivizing Reasoning Capability in LLMs for Multimodal Sequential Recommendation (MLLMRec-R1) — Hefei University of Technology, GRPO — arXiv 2603.06243 [opensource] - OT-FlowRec: Optimal Transport-Guided Multi-Bridge Flow for Multimodal Sequential Recommendation (OT-FlowRec) — Nanjing University of Science and Technology, arXiv 2610.16038 - MMWave: Tackling the Modality Inhibitance Effect with Wavelet-based Modality Integration in Multi-modal Recommendation (MMWave) — NTU, IEEE TKDE 2026 ### Optimizer - MuonRec - Robust Fusion of Semantic and Behavioural Signals for LLM Reranking in Personalised Search — Spotify; USRW Workshop @ RecSys 2026, arXiv 2609.25825 - CutBCE: A Fused Binary Cross-Entropy Loss for Large-Vocabulary Sequential Recommendation (CutBCE) - Google, arXiv 2610.05559 [opensource] ### Re-ranking - FARE: Deep Reinforcement Learning for Fair Exposure Constrained Uncertainty-Aware Financial Content Personalization (FARE) — JPMorganChase, arXiv 2609.31890 - The Recall Ceiling of LLM Recommendation Reranking (RecallCeiling) — University of Southern California, arXiv 2609.27953 - Bradley-Terry Rankings for Recommender Systems Across Dataset Taxonomies - Can LLM Rerankers Predict Their Own Ranking Performance? - CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring (CompRank) - Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking (CA-PG) - DeGRe: Dense-supervised Generative Reranking for Recommendation - EviRank: Evidence-Based Confidence Estimation for LLM-Based Ranking (EviRank) - GloRank (From Local Indices to Global Identifiers) - GR2 - Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance - GraphRAG-IRL (LLM re-ranking) - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - Joint Optimization of Relevance and Engagement in Multi-Task Ranking for E-Commerce with Efficient LLM Supervision - LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems - LRanker: LLM Ranker for Massive Candidates - miniReranker: Efficient Multimodal Reranking through Visual Cache Reuse and Interaction Sparsity - Ocean4Rec: Offline LLM-Derived OCEAN Profiles for Request-Time VOD Reranking - One Pass, Any Order: Position-Invariant Listwise Reranking for LLM-Based Recommendation (InvariRank) - OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System (OnePiece) - Profiling What Matters: Context-Aware Item Profiles from Large-Scale Metadata for LLM Recommenders (CAIRO) - Representation Curriculum: Stagewise Training for Robust Ranking and Allocation - ResRank - SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders (SORT) - Test-Time Training for Zero-Resource Dense Retrieval Reranking (DART) - Trading Engagement for Sustainability: Carbon-Aware Re-ranking for E-commerce Recommendations - UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale - Whole-Pool Setwise Reranking with Long-Context Language Models (WP-Setwise / DualEnd) - DrEM: Dual-Side Robust Ensemble Ranking from Noisy User Preference Predictions in Video Recommendation (DrEM) - SWIM: Step-Wise Integrated Measure for Session-supervised List Evaluation in Generative Re-ranking (SWIM) - Equal Ranking Quality, Different Decisions: Training Order-Consistent LLM Scorers - Beyond Polarization: The Generative Constraint of Chain-of-Thought in Pointwise Reranking - hLLM: Single Pass Decoding for Generative Reranking (hLLM) - DoPR: Reusable Compressed Document Prefixes for Efficient LLM Reranking (DoPR) - Embedding Surgery: Localized Updates for Adaptive Ranking Correction in Dense Retrieval - Recommendation Retrievers Need Verifiers: Universal Generative Reranking for Sequential Recommendations (Verifiers) - Balancing Trial and Reorder: A Hybrid Sequential Transformer-GBDT Ranker for On-Demand Delivery (UVR) — Wolt / DoorDash - Single-Token Expected-Value Scoring for Cold-Start Candidate Ranking - How Calibration Content Shapes Attention-Based Reranking (Interpolated Null Calibration) - Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text Reranking (MERIT-Rank) — Honor Device Co., Ltd, arXiv 2609.20131 - Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation (CRS-Performance) - KUAISHOU Explorer LLM-Rec Challenge 2026: Reasoning Generative Recommendation (OneRec-V2 / OneReason) — Kuaishou; SIGIR 2026 challenge; arXiv 2609.39828 - Routing Between Generative and Collaborative User Profiles: A Serving-Time Gate for Controllable Novelty — DePaul University / Comcast Technology AI; arXiv 2609.39043 - Decision-Oriented Recommendation Reranking: An Empirical Study of Jev (Jev) — University of Rochester / Meta AI, arXiv 2609.40241 ### Reasoning - From Interests to Semantic IDs: Retrieval-Grounded Credit Assignment for Generative Recommendation (Retrieval-Grounded Credit) - Learning Better Reasoning for Generative Recommendation with Semantic IDs (Evo-Rec) - A Reproducibility Analysis of PO4ISR: Diagnosing and Mitigating Semantic Drift in LLM-Based Session Recommendation - Agentic Recommender System with Hierarchical Belief-State Memory (MARS) - Bridging Behavior and Semantics for Time-aware Cross-Domain Sequential Recommendation (BST-CDSR) - Bridging Passive and Active: Enhancing Conversation Starter Recommendation via Active Expression Modeling (PA-Bridge) - Bringing Reasoning to Generative Recommendation Through the Lens of Cascaded Ranking (CARE) - Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking (CA-PG) - DeepInterestGR: Mining Deep Multi-Interest Using Multi-Modal LLMs - Discrimination Is Generation (DIG) - EviRank: Evidence-Based Confidence Estimation for LLM-Based Ranking (EviRank) - Expressiveness Limits of Autoregressive Semantic ID Generation in Generative Recommendation (Latte) - F-GRPO: Factorized Group-Relative Policy Optimization - Factorized Latent Reasoning for LLM-based Recommendation (FLR) - GPlan: Generative Spatiotemporal Intent Sequence Recommendation (GPlan) - GR2 - Generative Sequential Recommendation via Hierarchical Behavior Modeling (GAMER) - GraphRAG-IRL - Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation (ReaEmb) - HiLaR: Hierarchical Latent Reasoning for LLM-based Recommendation (HiLaR) - HoloRec: Holistic Encoding and Interleaved Reasoning for Generative Recommendation (HoloRec) - Implicit Reasoning for Large Language Model-based Generative Recommendation (PauseRec) - MixRAGRec: MoE KG-RAG for Multi-Agent LLM Recommendation - MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System (OnePiece) - OneReason Technical Report - OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework (OneSearch-V2) - PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation - RPORec: Reinforced Preference Optimization for Reasoning-Augmented Recommendations - ReRec: Reasoning-Augmented LLM-based Recommendation Assistant - Reasoning over Semantic IDs Enhances Generative Recommendation - RecRM-Bench: Benchmarking Multidimensional Reward Modeling - Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language (Restoring Collaborative) - Rethinking Convolutional Networks for Attribute-Aware Sequential Recommendation (ConvRec) - SAGER: Self-Evolving User Policy Skills for Recommendation Agent - Statistically Reliable LLM-Based Ranking Evaluation via Prediction-Powered Inference (PRECISE) - Stop Overthinking: Unlocking Efficient Listwise Reranking with Minimal Reasoning - ThinkGR: Integrating Chain-of-Thought into Generative Retrieval - TriAlignGR: Triangular Multitask Alignment with Multimodal Deep Interest Mining for Generative Recommendation - VRec (Verifiable Reasoning) - DREAMS: Dual-node Monte Carlo Tree Search for Conversational Recommendation (DREAMS) - Scaling Articulated Rationales for MLLM-based Recommendation (SARA) - Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding Learning (CoFree) — Alibaba (Taobao) / Wuhan University / Shanghai Jiao Tong University, arXiv 2609.20563 - Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text Reranking (MERIT-Rank) — Honor Device Co., Ltd, arXiv 2609.20131 - Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment (Re2A) — HK PolyU / CUHK / Sichuan University, EMNLP 2026 - Enhancing Group Recommendation with Memory-Augmented Reasoning in LLM Agent (AGR) - Iterative Semantic Reasoning from Individual to Group Interests for Generative Recommendation with LLMs (ISRF) - ReMem: Multi-memory GRPO for Long-Context Recommendation Agents (ReMem) - Hong Kong Polytechnic University / NTU, arXiv 2609.37311 - KUAISHOU Explorer LLM-Rec Challenge 2026: Reasoning Generative Recommendation (OneReason / OneRec-V2) — Kuaishou; SIGIR 2026 challenge; arXiv 2609.39828 - Beyond Sequences: Distilling Structured Decision Memory for LLM Recommendation (MARI) — Alibaba Group, arXiv 2610.11501 - OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models (OneLatent) — Kuaishou Technology, arXiv 2610.30675 ### Reinforcement Learning - FARE: Deep Reinforcement Learning for Fair Exposure Constrained Uncertainty-Aware Financial Content Personalization (FARE) — JPMorganChase, arXiv 2609.31890 - Learning Better Reasoning for Generative Recommendation with Semantic IDs (Evo-Rec) - Personalized and Trust-Aware Health Recommendation Policies for a Construction Workplace (Trust-Aware Health Rec) — University of Illinois Urbana-Champaign - EvoReason: Self-Evolving Reasoning Primitive-Guided On-Policy Distillation for Latent Reasoning in Generative Recommendation (EvoReason) - Exp-RSFT: Exponential Reward Weighting for Fine-Tuning Generative Recommenders under Sparse and Noisy Feedback (Exp-RSFT) - Progressive FM Post-Training: Progressive Alignment of Recommender Foundation Model through Multi-Phase Post-Training (Progressive FM Post-Training) - Feedback-Grounded Policy Discovery for Generative Recommendation (Feedback-Grounded) - GALA: Generative Aligned Learning for Adaptive Multimodal Representation (GALA) - HRPO: Hierarchical Residual Policy Optimization for Generative Recommendations (HRPO) - HiLaR: Hierarchical Latent Reasoning for LLM-based Recommendation (HiLaR) - RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems (RecHarness) - RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation (RecoReward) - RGD: Reward Guided Decoding for Generative Recommendation (RGD) - Think2Go: Generative Next POI Recommendation with LLM Reasoning (Think2Go) - Uncertainty-aware Generative Recommendation (UGR) - HCGRec: Hint-Conditioned Generative Recommendation with Semantic IDs (HCGRec) - Token-Level Credit Assignment Optimization for Generative Document Retrieval - Gwhere: Guess Where You Go — Generative Next Point-of-Interest Recommendation in Amap (Gwhere) - Astar: Learning to Propose Evolution Directions for Self-Evolving Industrial AI Systems (Astar) - ProRetrieval: Learning to Orchestrate Hybrid Search via Executable Program Synthesis (ProRetrieval) - It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning (CoGR) - World Model-Guided Reinforcement Learning via Counterfactual User Engagement Simulation (WMG-RL) - DMRL: Document-Mediated Reinforcement Learning for Skill Optimization in Advertising Recommendation (DMRL) - SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce (SAM-D2Q) - EAGER: Enrich-and-Align Generative Query Recommendation from Clicked Items in E-commerce Search (EAGER) - MemRetriever: Learning to Search, Reflect, and Retrieve from Long-Term Memory (MemRetriever) — GRPO - VARG: Value-Aware and Ranking-Aligned Generative Retrieval for Dynamic E-commerce Search (VARG) — Prefix-GRPO - Generate to Explore, Select to Exploit: Aligning LLM-based Headline Generation with Personalized Recommendation (GESE) — GSPO - Safety as a Constraint: Fine-Tuning a LLM Recommender to Explain Itself — constrained GRPO - TATK: Triple-Aware Top-K Learning with Knowledge-Grounded Verification for LLM-based Sequential Recommendation (TATK) — position-aware top-K rewards - LIGE-GR: A Smooth Leap from Ranking to Generative Recommendation in the LLM Era (LIGE-GR) - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 - Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding Learning (CoFree) — Alibaba (Taobao) / Wuhan University / Shanghai Jiao Tong University, arXiv 2609.20563 - PAPA: Online Personalized Active Preference Alignment (PAPA) — Washington University in St. Louis, arXiv 2607.00486 - Enhancing Group Recommendation with Memory-Augmented Reasoning in LLM Agent (AGR) - ReFilter: Bridging Embeddings and LLM Filtering for Similar Mobile App Retrieval (ReFilter) — University of Toronto / Université du Québec à Montréal; ASIS&T 2026, arXiv 2609.25306 - DASO / Difficulty-Aware Semantic-ID Optimization (GRPO rollout-allocation) — Meta / Penn State — [Also published on 2026-09-23] - GRP: Snap's Generative Recommendation Paradigm with mGRPO Reward-Guided Post-Training (GRP) - Snap Inc., arXiv 2609.36688 - ReMem: Multi-memory GRPO for Long-Context Recommendation Agents (ReMem) - Hong Kong Polytechnic University / NTU, arXiv 2609.37311 - Reinforcement Learning with Conformal Action Sets: An Application to Sequential Recommendation (RLCP) — University of Pennsylvania, arXiv 2610.08743 - Personal-Agent Mediated Recommendation with Cross-Platform User History (MediateRec) — UC San Diego / Meta (PAMO policy-optimization mediation), arXiv 2610.07588 - MLLMRec-R1: Incentivizing Reasoning Capability in LLMs for Multimodal Sequential Recommendation (MLLMRec-R1) — Hefei University of Technology, GRPO — arXiv 2603.06243 [opensource] - SafeCRS: Personalized Safety Alignment for LLM-Based Conversational Recommender Systems (SafeCRS) — UIC / UIUC / Amazon, Safe-GDPO, KDD 2026 — arXiv 2603.03536 [opensource] - RAR: Retrieval Augmented Conversational Recommendation with Reinforcement Learning (RAR) — UIUC / Google DeepMind / UC San Diego — arXiv 2604.04457 [opensource] - Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders (ReMaRk) — Nanyang Technological University, arXiv 2610.17771 [opensource] - SemPOI-RL: Aligning LLM Semantic Reasoning for Interpretable Out-of-Town POI Sequential Generation (SemPOI-RL) — Beihang University, arXiv 2608.30399 [opensource] ### Retrieval / Representation - ARMED: Adaptive Retrieval-Augmented Medication Recommendation (ARMED) — UX Factory / KAIST, ICTAI 2026, arXiv 2610.16440 - CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring (CompRank) - Tail-Aware Adaptive-k: Query-Adaptive Context Selection for Retrieval-Augmented Generation (TAA-k) - Training-Free LLM-Based Recommendation with Post-LLM Item Refinement Using Collaborative Signals (CoRRe) - What Limits Does Quantization Place on Dense Top-k Retrieval? A Theoretical Study - Beyond Patches: Superpixel Token-based Transformers for Attribute-Specific Fashion Retrieval (SuperFashion) - Bridging Short Videos and Live Streams: Reasoning-Guided Multimodal LLMs for Cross-Domain Representation Learning (RGCD-Rep) - Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking (CA-PG) - Distributional Approximate Nearest Neighbour Search for Uncertainty-Aware Retrieval (DINOSAUR) - Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors - Efficient Generative Retrieval for E-commerce Search with Semantic Cluster IDs and Expert-Guided RL - Generalizing Graph Foundation Models via Hyperbolic Retrieval-Augmented Generation (HyRAG) - INSPIRE: Intent-aware Neural Sponsored Product Retrieval for E-commerce - Learning User-Aware Recall: Personalized Retrieval in Long-Term Conversational Memory (PPRO) - LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation - MARC (LLM representation compression) - Memento: Personalized RAG-Style Long-Retention Data Scaling for META Ads Recommendation - MixRAGRec: MoE KG-RAG for Multi-Agent LLM Recommendation - Attention Calibration for Position-Fair Dense Information Retrieval - RAGEAR: Retrieval-Augmented Graph-Enhanced Academic Recommender - Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval - ResRank (retrieval + reranking) - Skill Is Not Document: A Query-Conditional Benchmark and Two-Stage Retriever for LLM Agent Skill Routing (R3) - The Case Against Generation for Retrieval: Discriminative Language Models as Effective Retrievers (Discriminative Retrieval) - Tokenizing Numerical and Embedding Features for LLM RecSys - Serving the Long Tail: Training-Free LLM Candidate Generation for Vacation Rental Marketplaces - Scaling Dense Retrieval with LLM-Annotated Training Data: Structured Mining and Progressive Curriculum for E-Commerce Sponsored Search - STORM: Stepwise Token Optimization with Reward-Guided Beam Search - ICICLE: Expanding Retrieval with In-Context Documents (ICICLE) - TransRetrieval: Scaling Up Transformer-Based Retrieval for Industrial Recommendation (TransRetrieval) - Embedding Surgery: Localized Updates for Adaptive Ranking Correction in Dense Retrieval - SAGE: Semantic Attribute Graphs for Multi-Entity Visual Retrieval (SAGE) - Inventory-Grounded Policy-Level Optimization for Training-Free AI Search (IGPO) - Generative Late-Interaction Embeddings For Visual Document Retrieval (GLIE) - Matryoshka Hash Representations for Model-Aware Compact Semantic Retrieval (MHR) - Query-Aware Token Budgeting for Efficient Late-Interaction Visual Document Retrieval - RegRet: Enhancing Region-Level Retrieval in Large Multimodal Models (RegRet) - Efficient Swing Computation for Retrieval in Large-Scale Recommender Systems (ASC / K-ASC) - How Calibration Content Shapes Attention-Based Reranking (Interpolated Null Calibration) - Quanta: A Self-Contained Python Library for Hybrid Retrieval over Quantised Embeddings, Lexical Indexes, and Knowledge Graphs (Quanta) - Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding Learning (CoFree) — Alibaba (Taobao) / Wuhan University / Shanghai Jiao Tong University, arXiv 2609.20563 - Self-Evolving Search Index (SELF-INDEX) — Yonsei University, arXiv 2609.19656 - Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text Reranking (MERIT-Rank) — Honor Device Co., Ltd, arXiv 2609.20131 - Not All Is Lost: Repairing Lossy User Preference States of Personalization Encoders (REPAIR) — IIIT Delhi, arXiv 2610.01270 - RAR: Retrieval Augmented Conversational Recommendation with Reinforcement Learning (RAR) — UIUC / Google DeepMind / UC San Diego, retrieval-augmented generative CRS — arXiv 2604.04457 [opensource] ### Scaling Laws - Farewell to Item IDs: Unlocking the Scaling Potential of Large Ranking Models via Semantic Tokens (TRM) - Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation (Scaling Laws) - RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems (RankUp) - SSR: Beyond Dense Connectivity: Explicit Sparsity for Scalable Recommendation (SSR) - Towards Generalizable and Efficient Large-Scale Generative Recommenders (Netflix GenRec Scaling) - SITA: Semantic Interest Tokens for Target-Aware Compression in Long-Sequence Recommendation (SITA) - SmartGR: Hierarchy and Beam-Aware Knowledge Distillation for Generative Recommendation (SmartGR) - UnpairGR: Unpaired Modality-Agnostic Generative Recommendation (UnpairGR) - Understanding Generative Recommendation with Semantic IDs from a Model-scaling View (SID Scaling) - UniMixer: A Unified Architecture for Scaling Laws in Recommendation Systems (UniMixer) - TransRetrieval: Scaling Up Transformer-Based Retrieval for Industrial Recommendation (TransRetrieval) - From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs (ReST) - From a Static Multi-Level Small Semantic Codebook to a Dynamic Single-Level Large Semantic Codebook for Generative Recommendation — Kuaishou, arXiv 2608.21012 - HELIX: Rethinking Scaling Laws for Large-Scale Recommendation (HELIX) - TikTok, arXiv 2609.37183 - KUAISHOU Explorer LLM-Rec Challenge 2026: Reasoning Generative Recommendation (OneReason / OneRec-V2) — Kuaishou; OneReason uses advanced RL to make CoT reasoning beneficial; SIGIR 2026 challenge; arXiv 2609.39828 - LoopGR: Parameter-Efficient Scaling for Generative Recommendation via Recurrent Depth (LoopGR) — Tencent, arXiv 2610.34157 ### Semantic / Structured IDs - ReGen: Controllable Generative Recommendation via Guided Token Refinement (ReGen) — XJTLU / U Liverpool / PolyU, arXiv 2610.28642 [opensource] - FineSID: Progressive Learning from Content to Relevance to Personalization for CTR Prediction (FineSID) — Alibaba, arXiv 2610.29138 - SPAR: Stepwise Personalized Autoregressive Recommendation with Prefix-Balanced Semantic IDs (SPAR) — UESTC & Alibaba, arXiv 2610.27898 - Revisiting Generative vs. Dense Retrieval: A Static Value Prior Reverses the Gap (Static Value Prior) — Shopify, arXiv 2610.19065 - From Interests to Semantic IDs: Retrieval-Grounded Credit Assignment for Generative Recommendation (Retrieval-Grounded Credit) - Learning Better Reasoning for Generative Recommendation with Semantic IDs (Evo-Rec) - CogRec: Structure-Cognitive Fast-and-Slow Reasoning for Generative Recommendation (CogRec) - Once Generated, Ranked: End-to-End Generative Slate Recommendation with Unified Semantic-Collaborative IDs (OGR) - OxygenREC-v2: Internalizing Discrimination into Generative Recommendation (OxygenREC-v2) - PaletteID: Prototype-Composed Semantic Identifiers for Multimodal CTR Prediction (PaletteID) - SIDScope: A Diagnostic Resource for Semantic-ID Interfaces in Generative Recommendation (SIDScope) - SnapLGR: LLM-Based Generative Retrieval for Snapchat Content Recommendation (SnapLGR) - SSR-GRPO: Integrating Supervision and Semantic IDs into Reinforcement Learning for Dense Retrieval in E-commerce (SSR-GRPO) - Think2Go: Generative Next POI Recommendation with LLM Reasoning (Think2Go) - ACE: Anisotropy-Controllable Embedding for LLM-enhanced Sequential Recommendation - Beyond Instance-Level Alignment and Uniformity: Semantic Factor Learning for Collaborative Filtering (SaFeAU) - Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling - Beyond Matching: Category-Guided Latent Intent Reasoning for Generative Retrieval in E-Commerce (CaLIR) - Beyond Static Collision Handling: Adaptive Semantic ID Learning (AdaSID) - Beyond Uniform Token Training: A Multi-Target Framework for Learning Token-Weighted Objectives in Generative Recommenders (Beyond Uniform Token Training) - CARD: Non-Uniform Quantization of Visual Semantic Unit for Generative Recommendation - Causal Representation Learning for Generalisable Recommendation - ChronoID: Infusing Explicit Temporal Signals into Semantic IDs for Generative Recommendation (ChronoID) - Closing the Indexing-Decoding Gap in Multimodal Generative Retrieval via Prefix Retention Optimization (PRO) - Conditional Memory Enhanced Item Representation for Generative Recommendation (ComeIR) - Decoupled Residual Quantization for Robust Semantic IDs in Recommendation (DRQ) - DECOR: Learning Decomposed Contextual Token Representations from Pretrained and Collaborative Signals for Generative Recommendation (DECOR) - Deep Interest Mining (SemanticID) - Discrimination Is Generation: Unifying Ranking and Retrieval from a Tokenizer Perspective (DIG) - DOS: Dual-Flow Orthogonal Semantic IDs for Recommendation in Meituan (DOS) - DSIRM: Learning Query-Bridged Discrete Semantic Identifiers for E-commerce Relevance Modeling (DSIRM) - Expressiveness Limits of Autoregressive Semantic ID Generation in Generative Recommendation (Latte) - Farewell to Item IDs: Unlocking the Scaling Potential of Large Ranking Models via Semantic Tokens (TRM) - FusID: Modality-Fused Semantic IDs for Generative Music Recommendation (FusID) - Generative Sequential Recommendation via Hierarchical Behavior Modeling (GAMER) - GLIDE: Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify (GLIDE) - GLASS: Coarse-to-Fine Long-term Interest Modeling for Generative Recommendation (GLASS) - GRACE: Generative Recommender Acceleration Engine for Real-Time Ads Retrieval (GRACE) - Grevo: A Unified Generative Recommendation Framework with Evolutionary Item Indexing (Grevo) - GRLM - Grounded Token Initialization for New Vocabulary in LMs for Generative Recommendation (GTI) - Gryphon: A Unified Architecture for Semantic-ID Generation and Item-Level Scoring in Industrial Recommendations (Gryphon) - Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation (ReaEmb) - HRPO: Hierarchical Residual Policy Optimization for Generative Recommendations (HRPO) - Hyperbolic RQ-VAE enhanced Generative Recommendation with Differential-Length Codebook Strategy (HG-Rec) - How Reliable Are Semantic-ID Tokenizer Comparisons in Generative Recommendation? - Implicit Reasoning for Large Language Model-based Generative Recommendation (PauseRec) - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation - IntRR: A Framework for Integrating SID Redistribution and Length Reduction - LLMs Need Encoders for Semantic IDs Too (PrefixMem) - MERGE: Next-Generation Item Indexing Paradigm for Large-Scale Streaming Recommendation (MERGE) - Mitigating Collaborative Semantic ID Staleness in Generative Retrieval - On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (TA-Rec) - One Hierarchy, Two Systems: Semantic Product IDs for Discovery-Surface Ranking and Search-Page Query Reformulation - One Pool, Two Caches: Adaptive HBM Partitioning for Accelerating Generative Recommender Serving - PIT: A Dynamic Personalized Item Tokenizer for End-to-End Generative Recommendation (PIT) - PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation - Quantizing Intent: Cross-Domain Semantic IDs from Organic Activity for Industrial Ranking - Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale (RaG) - Reasoning over Semantic IDs Enhances Generative Recommendation - Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language (Restoring Collaborative) - RGD: Reward Guided Decoding for Generative Recommendation (RGD) - SAILRec: Steering LLM Attention to Dual-Side Semantically Aligned Collaborative Embeddings for Recommendation (SAILRec) - SPARC: Sequence-aware Progressive Attribute Routing and Compression Framework for Generative Recommendation (SPARC) - SIDInspector: A Mapping-First Diagnostic Resource for Semantic-ID Tokenizers - SIREN: Unified Multi-Granularity Semantic Interaction for Multi-Modal Lifelong User Interest Modeling - ShopX: A Foundation Model for Intent-to-Item Fulfillment in Agentic Shopping (ShopX) - SSRLive: Live Streaming Recommendation with Dynamic Semantic ID (SSRLive) - Stop Treating Collisions Equally: Qualification-Aware Semantic ID Learning for Recommendation at Industrial Scale (QuaSID) - Structuring and Tokenizing Distributed User Interest Context for Generative Recommendation (G2Rec) - TCA4Rec: Token-level Collaborative Alignment for LLM-based Generative Recommendation - TopoGR: Revealing and Preserving Latent Structure of Semantic ID in Generative Recommendation (TopoGR) - TriAlignGR: Triangular Multitask Alignment with Multimodal Deep Interest Mining for Generative Recommendation - TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems (TokenMinds) - Tokens are All You Need: Dual-purpose Semantic IDs for Achieving LLM-Level I/O Efficiency in Recommendation Systems - TwiSTAR: Think Fast, Think Slow, Then Act, Generative Recommendation with Adaptive Reasoning - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - UniRec: Bridging the Expressive Gap via Chain-of-Attribute - UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence - VaLiDRec: Variable-Length LLM-Aligned Semantic IDs for Generative Recommendation (VaLiDRec) - When Text-as-Vision Meets Semantic IDs in Generative Recommendation: An Empirical Study - DREAM: Dynamic Refinement of Early Assignment Mappings (DREAM) - Do Generative Recommenders Deepen the Information Cocoon? A Closed-Loop Simulation with LLM-powered User Simulators (RecLoop) - The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation (H2Rec) - Beyond Semantic IDs: Encoding Business-Value Ranking into Document Identifiers for Generative Retrieval (CRID) - IBA: Where Reasoning Matters (IBA) - Interpretable Representation via LLM-Driven Generative Disentanglement for Local-Life Service Recommendation (LGRID) - Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation (Tencent/ACL 2026) - MLPs are Efficient Distilled Generative Recommenders (SID-MLP) - TwiSTAR: Think Fast, Think Slow, Then Act — Generative Recommendation with Adaptive Reasoning (TwiSTAR) - HCGRec: Hint-Conditioned Generative Recommendation with Semantic IDs (HCGRec) - FSGR: Mitigating Token Frequency Bias for Fair SID-Based Generative Recommendation (FSGR) - Gwhere: Guess Where You Go — Generative Next Point-of-Interest Recommendation in Amap (Gwhere) - EchoRec: Multi-Item Prediction-Empowered Generative Recommendation via Cycle-Consistent Preference Alignment (EchoRec) - Rethinking Item Tokenization in Generative Recommenders: From Fixed Atoms to Semantic Subwords (SST) - PRQ-KMeans: Projection Residual Quantization for Semantic ID Tokenization (PRQ-KMeans) - Tlow: Flow-based Item Tokenizer for Recommendation (Tlow) - An Event is Worth One Token: Event Tokenization for Industrial-scale LLM Recommendation (AMBER) - CoFiRec: Coarse-to-Fine Tokenization for Generative Recommendation (CoFiRec) - HypRQ-VAE: Hyperbolic Item Indexing for Long-Tail-Aware Generative Recommender Systems (HypRQ-VAE) - HF-SID: High-Fidelity Semantic IDs for Generative Retrieval in Location-Based Services (HF-SID) - Exploring Bottom-Up Clustering for Creating Semantic IDs - REDSI: Addressing the Reproducibility and Evaluation Consistency of Differentiable Search Indexing for Document Retrieval (REDSI) - PDMR: Passage-Driven Multi-ID Document Retrieval (PDMR) - Addressing Cross-Stage Decoupling of Semantic and Collaborative Signals in Generative Recommendation (SCRec) - ReliGRec: Reliability-Oriented LLM-Based Generative Recommendation via User-Risk-Aware Prompt Routing (ReliGRec) - One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations (ANGLE) - UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising (UniPolicy) — Meituan, arXiv 2609.20630 - Attribute-Conditioned Multimodal Slot Factorization for Controllable Fashion Retrieval (MM-slotgate) — Amazon, arXiv 2608.12570 - Drift-Aware Continual Tokenization for Generative Recommendation (DACT) - Iterative Semantic Reasoning from Individual to Group Interests for Generative Recommendation with LLMs (ISRF) - What Makes a Good Semantic ID for Generative Recommendation? A Reproducibility Study (SID-Repro) — Shandong U / Glasgow / Leiden; SIGIR-AP 2026; large-scale SID-design reproducibility study - Guiding the coarse levels of semantic IDs makes the fine levels learnable (Guided SID) — Meta; deterministic supervised index assignment pins coarse RQ-VAE levels to text-grounded attributes - SpeakGR: Preserving Language Generation While Learning Semantic IDs (SpeakGR) - Imperial College London, arXiv 2609.35430 - FineSID: Scalable and Efficient Semantic Identifier Learning for Generative Recommendation — Tsinghua University / Huawei Noah's Ark Lab / USTC; arXiv 2609.36670 - Residual Trajectory Distillation for Generative Retrieval (ResTD) — Beihang University / Meituan; arXiv 2609.39319 - Learning Multiresolution Relevance for Hierarchical Generative Retrieval (RARS) — Beihang University / Meituan; arXiv 2609.39312 [opensource] - GEAR: Generative End-to-end Ad Retrieval at Douyin (GEAR) — ByteDance (Douyin); arXiv 2609.39327 - Graph-Informed Semantic IDs (GrIS): Balancing Semantic and Collaborative Signals via Recursive Graph Partition (GrIS) — Huawei Ireland Research Centre, arXiv 2610.01533 - Do Multilingual Encoders Produce Language-Consistent Semantic IDs? (MultiLing-SID) — Amazon / Rutgers, arXiv 2610.01139 - SPRIG: Semantic-ID-enhanced Paths for Knowledge Graph-based Generative Recommendation (SPRIG) — Johannes Kepler University Linz, CIKM 2026, arXiv 2610.06590 [opensource] - A Systematic Study of Semantic ID Spaces for Generative Information Retrieval (Semantic-ID Spaces) — Artefact Research Center / Université d'Angers (LERIA), arXiv 2610.08732 - FLASH: Rethinking Semantic ID Construction for Generative Recommendation: SimHash with Parallel Decoding and Semantic Alignment (FLASH) — U Illinois Chicago / NeurIPS 2026, arXiv 2610.07402 - DiffGRM: Diffusion-based Generative Recommendation Model (DiffGRM) — Kuaishou, WWW 2026 (replaces AR decoder with masked discrete diffusion) — arXiv 2510.21805 [opensource] - Bit-Factorized: Bit-Factorized Identifiers for Efficient Generative Retrieval (Bit-Factorized) — Leiden University / Radboud University, arXiv 2610.17467 - HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation (HASTE) — Glance, arXiv 2610.17368 - TacGR (UniSGR-V2): Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation (TacGR) — Alibaba International, arXiv 2610.16132 - Personalized LLM Recommendation with Semantic IDs: Autoregressive Beam Search vs. Multi-Token Prediction (Netflix SID Study) — Netflix, NeurIPS 2026 PALM Workshop, arXiv 2610.15864 - LLM-based User-Product Relevance Evaluation for Dynamic Product Advertising (DPA Relevance) — Snap Inc., CIKM 2026, arXiv 2610.14550 - AGR: Agentic Generative Retrieval with Tool-Augmented Search (AGR) — Kuaishou Technology & Fudan University, arXiv 2610.26831 - Don't Rewrite the SID Map: Topology Preservation for Generative Retrieval in Dynamic E-Commerce Search (SID-Map) — Alibaba, arXiv 2610.26105 - Structure-Aligned History Learning with Subspace-Structured Semantic IDs for Generative Recommendation (SAHL) — Chongqing University, arXiv 2610.25093 - OneChord: Distill What You Rank, Recommend in One Pass (OneChord) — University of Chinese Academy of Sciences, arXiv 2610.25805 - OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models (OneLatent) — Kuaishou Technology, arXiv 2610.30675 ### Security / Adversarial - Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems — Pinterest, arXiv 2609.33073 - CoSimRec: Measuring Coordinated-Content Penetration in Recommender Feedback Loops (CoSimRec) - VENOMREC: Cross-Modal Interactive Poisoning for Targeted Promotion in Multimodal LLM Recommender Systems (VENOMREC) - Fairness Attacks on Recommender Systems - SIREN (Luring LLMs onto the Rocks): PAIR-Driven Preference Manipulation in Web-RAG Recommenders (SIREN) - An Efficient and Effective Agentic Group Shilling Attack on Recommender Systems (AGAS) - The Like Trap: Multi-Stage Poisoning against Agents in Similarity-based Recommendation Systems (Like Trap) — Michigan State University, arXiv 2609.27155 - ReliGRec: Reliability-Oriented LLM-Based Generative Recommendation via User-Risk-Aware Prompt Routing (ReliGRec) — weak-risk shilling-robust prompt routing - VirusCascade: Hijacking Collaborative Reflection in LLM-Powered Recommender Agents — Nanyang Technological University, arXiv 2609.38270 - AlignPoison: Conditional Targeted Poisoning of Generative Recommenders via Decoding-Aligned Prefixes (AlignPoison) — UNC Chapel Hill / Notre Dame / Meta, arXiv 2610.14487 ### Sequential Modeling / RoPE - LSF-SR: Latent Semantic Fusion for Sequential Recommendation via Flow-based Conditional Variational Autoencoders (LSF-SR) — National Yang Ming Chiao Tung University, arXiv 2609.29815 - Learning from the Future: Privileged Self-Distillation for Sequential Recommendation (PSD) - MIRAGE: Escaping the Euclidean Void (Manifold-Informed Flow Matching for Sequential Recommendation) - SITA: Semantic Interest Tokens for Target-Aware Compression in Long-Sequence Recommendation (SITA) - A Rank-One Popularity Component in Dot-Product Recommender Scores: Population Theory and Prior-Separation Evidence - ACE: Anisotropy-Controllable Embedding for LLM-enhanced Sequential Recommendation - APAO: Bridging the Training-Inference Gap in Generative Recommendation via Adaptive Prefix-Aware Optimization (APAO) - BAHSD: Bridging the Long-tail Gap via Adaptive Distillation in Black-box Sequential Recommendation (BAHSD) - Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling - ChronoID: Infusing Explicit Temporal Signals into Semantic IDs for Generative Recommendation (ChronoID) - CMSL: Constructive Multi-Sequence Learning for Recommendation Systems (CMSL) - Divergence Meets Consensus: A Multi-Source Negative Sampling Framework for Sequential Recommendation (MDCNS) - FAVE: Flow-based Average Velocity Establishment for Sequential Recommendation - FOSTER: First-order Dataset Distillation for Text-based Sequential Recommendation - Preference-Drift-Aware Subsequence Learning and Hierarchical Context Fusion for Long-Sequence Generative Recommendation - ChronicleRec: Pre-training Temporally Anchored Tokens for Lifelong User Modeling (ChronicleRec) - Generative Archetype-Grounded Item Representations for Sequential Recommendation (GenAIR) - GPlan: Generative Spatiotemporal Intent Sequence Recommendation (GPlan) - Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation (ReaEmb) - HoloRec: Holistic Encoding and Interleaved Reasoning for Generative Recommendation (HoloRec) - IAT: Instance-As-Token Compression for Historical User Sequence Modeling (IAT) - IID-Nav: From Extraction to Navigation: Progressive Retrieval with Indirectly Infinite Depth (IID-Nav) - Learning to Forget: Satiation-Aware Long-Sequence Transducers for Mitigating Post-Purchase Redundancy (SAM) - LoopMemGR: From Behavior Logs to Evolving Memory for Generative Recommendation (LoopMemGR) - MARS: Multi-rate Aggregation of Recency Signals for Sequential Recommendation across Sparse and Dense Regimes (MARS) - One Sequential Recommendation Model Pretrained from Synthetic Priors Predicts Multiple Datasets (SRPFN) - PHKT: Personalized Dynamic Hypergraph-enhanced KAN-Transformer for Multi-behavior Sequential Recommendation (PHKT) - SAERec: Constructing Fine-grained Interpretable Intents Priors via Sparse Autoencoders for Recommendation (SAERec) - Segmentation-Supervised Complexity-Adaptive Recommendation (S2-CAR) - SIREN-RoPE (Learning to Rotate) - SIREN: Unified Multi-Granularity Semantic Interaction for Multi-Modal Lifelong User Interest Modeling - SPARC: Sequence-aware Progressive Attribute Routing and Compression Framework for Generative Recommendation (SPARC) - SSRLive: Live Streaming Recommendation with Dynamic Semantic ID (SSRLive) - TAP-PER: Beyond Retrieval: Learning Compact User Representations for Scalable LLM Personalization (TAP-PER) - The Pitfall of Scaling Up: Uncovering and Mitigating Popularity Bias Amplification in Scaling Transformer-based Recommenders (SPRINT) - IMFuse: Instance-Aware Multi-Layer Fusion for LLM-Enhanced Sequential Recommendation (IMFuse) - Time-Aware Diffusion based on Preference Disentanglement for Generative Recommendation (TDPM) - UFRec: Uncertainty-Guided Future Learning for Sequential Recommendation - UniDot: A Unified Network for Sequence Modeling and Feature Interaction in Large-scale Recommendation (UniDot) - UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation (UniFormer) - UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence - VirtualMLE: A Virtual ML Engineer that Optimizes Sequential Recommenders (VirtualMLE) - TRUST: Item-Calibrated Interval Evidence for Temporal Session-Based Recommendation (TRUST) - RecRec: Recursive Refinement for Sequential Recommendation - RecRec: Latent Interests Recursive Reasoning for Sequential Recommendation (RecRec) - Stream-aware Side Adaptation for Large Pre-trained Multimodal Embedding Models in Sequential Recommendation (Stresa) - The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation (H2Rec) - Making Collaborative Signals Count: Graph-Aware Large Language Models for Sequential Recommendation (GALLM) - From Overlooked to Explored: Recovering Item Relations via Mixture of Perspectives for Sequential Recommendation (PRISM) - STAR: Structured Tokenization and Target-Aware Interest Representation for PCVR Prediction (STAR) - Decoupled Temporal Encoding for Generative Recommendation (DTE) - GOD: Enhancing Generalization via Deep Grafting for Sequential Recommendation (GOD) - SAGA: Structure-Attended Generative Action Embedding Model that encodes Multi-Surface User Action Sequences (SAGA) - Residual Dominance as a Structural Account of Last-Item Reliance in Causal Self-Attention Recommenders (Residual Dominance) - Hamiltonian Spectral-Temporal Dissipative Dynamics for Sequential Recommendation (HSR) - SG-UMP: Sequence-Guided Universal Multimodal Prioritization Calculation Framework (SG-UMP) - SwapRec: Warming Up Cold Items Through Training-Time Swaps (SwapRec) - Recommender System as Slow and Fast Thinkers (DS-Frame) - Two-Sided State-Space Models for Sequential Recommendation with Non-Random Multimodal Review Feedback (TS-SSM) - SequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching (SequenceO1) - Closing the Long-Short View Gap in Sequential Recommendation without Cached History - Balancing Trial and Reorder: A Hybrid Sequential Transformer-GBDT Ranker for On-Demand Delivery (UVR) - SURF: Subtractive Updates for Recommender Forgetting (SURF) - Dense Feature Representation over Sequence Modeling: A Solution to the KDD Cup 2026 UniRec Challenge (KDD Cup 2026 UniRec) — Z Lab (Chengdu), KDD Cup 2026 UniRec Workshop - Give the Long-tail More SPACE: Promoting Provider Fairness in Next POI Recommendation (SPACE) — Southeast University, arXiv 2608.07998 - Drift-Aware Continual Tokenization for Generative Recommendation (DACT) - Beyond Interleaving: Causal Attention Reformulations for Generative Recommender Systems - MuSeR: Scalable Long-sequence Recommendation with Multi-interest Modeling (MuSeR) — Baidu / CityU HK / CUHK; hierarchical temporal compression + multi-interest + beam-search retrieval, deployed at Baidu - A Redundancy Reduction Approach for Controllable Sequential Recommendations (BT-SR) — Yandex / AIRI / HSE; Barlow-Twins decorrelation for head/tail exposure control - CRAMER: Control via Request-Aware Masking for Editing Recommenders (CRAMER) — Renmin University of China / Dalhousie University; arXiv 2608.25370 - T-RoPE: Time-Aware Rotary Position Embedding for Sequential Recommendation (T-RoPE) — Shopify, arXiv 2609.30576 - Recommendation World Models for Future-State Control (UA-TWM) — University of British Columbia et al., arXiv 2609.30711 - RouteRec: Behavior-Guided Sparse Routing for Sequential Recommendation (RouteRec) — KAIST / Seoul National University, arXiv 2609.39007 [opensource] - Learning Robust Personalized Prompts for LLM-Driven Sequential Recommendation (LRPRec) — Zhejiang University, arXiv 2610.03923 - Reinforcement Learning with Conformal Action Sets: An Application to Sequential Recommendation (RLCP) — University of Pennsylvania, arXiv 2610.08743 - FatsMB: From Agnostic to Specific: Latent Preference Diffusion for Multi-Behavior Sequential Recommendation (FatsMB) — CAS / Kuaishou, arXiv 2602.23132 - DiffSBR: Diffusion-based Latent Neighbor Generation for Session-based Recommendation (DiffSBR) — UESTC, arXiv 2601.03903 - DiffuReason: Bridging Latent Reasoning and Generative Refinement for Sequential Recommendation (DiffuReason) — Tencent, arXiv 2602.09744 - Multimodal Graph Retrieval-Augmented Sequential Recommendation via Collaborative Filtering Paths (MGRASRec) — University of New South Wales, arXiv 2610.11228 - CARD: Counterfactual Attention Regulation Diffusion for Diffusion-based Sequential Recommendation (CARD) — Sun Yat-sen University, ICASSP 2026 — arXiv 2601.15673 [opensource] - PR-entmax: Mitigating Popularity Bias in Sequential Recommendations with Sparse Prior-Regularized Objective (PR-entmax) — Sber AI Lab / Lomonosov MSU, CIKM 2026 — arXiv 2610.17823 - BGAFlow: Bezier-Guided Adaptive Flow for Cross-domain Sequential Recommendation (BGAFlow) — Anhui University, arXiv 2610.30825 - Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation (LoopRec) — Renmin University of China, arXiv 2610.30128 [opensource] ### Survey - Rethinking Fairness in LLM-Based Recommender Systems: A Survey - Trustworthy Recommendation in the Era of Large Language Models: Opportunities and Challenges - Behavior-Mining, Generative Conversations, and Collaborative Advisory: the Future of Travel and Tourism Recommender Systems (TTRS) — UAM Madrid / TU Munich / Free University of Bozen-Bolzano, arXiv 2610.08232 ### Temporal Stability / Feature Pruning - Decomposing Staleness in Recommender Systems: A Dual-Filter Framework for Supersession and Decay (SDF) - Fortress: A Case Study in Stabilizing Search Recommendations ### Training Systems / NPU Optimization - Not Only NTP: Extending Training Signal Coverage for Generative Recommendation (NONTP) — Meituan, arXiv 2607.12277 - TurboGR: An Accelerated Training System for Large-Scale Generative Recommendation