Paper catalog

English home · 中文研究库

The catalog lists 1061 entries from the shared research manifest. English write-ups available: 1061.

Browse by domain

• Search, ads, recommendation, and LLM applications: 324 catalog entries, 324 English write-ups
• Foundation models: 204 catalog entries, 204 English write-ups
• Structured decisions / System One: 6 catalog entries, 6 English write-ups
• Multimodal models: 17 catalog entries, 17 English write-ups
• LLM post-training: 244 catalog entries, 244 English write-ups
• Agents: 266 catalog entries, 266 English write-ups

Read in English

• Strategist: Evolving Reasoning with Agents
• SpecOPD: Branch-Parallel On-Policy Distillation for Speculative Decoding
• RLiD: Reinforcement Learning in the Decision Plane for Language Agents and LLMs
• XLA from the Trenches: Correctness on TPU Workloads at Scale
• WeSearch: A Unified Model-Centric Retrieval Framework for WeChat Search
• ProLand: A New Infeasible Manifold Optimization Method for LoRA Fine-Tuning
• GAVAGAI: Referential Alignment for Infant-Scale Vision-Language Models via Balanced Optimal Transport
• AgentHeal: Evaluating Goal-Preserving Resource Recovery of Web Agents under Access Barriers
• Where Should the Evaluation Budget Go in LLM-Driven Agent Scaffold Evolution?
• When the Critic Falls Behind: Co-Evolving Rubrics and Harnesses for Continual Agent Self-Improvement
• Select-TTT: Query-Conditioned Selective Fast Weight Updates for In-Place Test-Time Training
• SWIM: Learning What Matters from Behaviors for JEPA World Models
• Remembering Is Not Enough: Refining Memories and Learning When to Use Them in Multi-Agent Systems
• Proposing Better Rollouts for On-Policy Distillation
• PailitaoMMSearch2.0 Technical Report: Scaling Generative Retrieval to Unified End-to-End Multimodal Search
• Multi-State Linear Attention with Generalized Second-Order Routing
• KL-MOPD: Discrepancy-Aware Multi-Objective On-Policy Distillation for Generative Recommendation
• GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents
• Folding Attention Biases into Keys Makes Compacted KV Caches Run on Standard Kernels
• CONCUR: Concordant Auxiliary Rewards for Multi-Task, Multi-Reward Reinforcement Learning
• ResCom: Activation Residual Compensation for Large Language Model Quantization
• EMMA: Structured Episodic Memory for Multi-Agent Workflow Synthesis
• PinGER: Token-Efficient Generative Retrieval with Semantic IDs
• When to Stop Penalizing Actions? Martingale GRPO for VLA Reinforcement Learning
• UniAction: Unified Action Conditioning for World Models
• Reinforcement Learning for Controller Parameters in the Auto-Bidding System at Pinterest
• RSI-Opt: Recursive Self-Improvement Optimizer Agents For Model Training
• PathJev: Decision-Native Vision–Language Modelling for Histopathology
• On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models
• ImageRSI: Verification-Driven Recursive Self-Improvement for Agentic Image Generation
• Hierarchical GTV Estimation: Bridging the Gap between Container-Level Ranking and Item-Level Conversion
• HarnessMM: Recursive Self-Improvement for Harness Token Efficiency
• Frontier-OPD: Discovering the Edge of Agent Capability and Teaching Agents to Go Beyond
• FlexGlue: Sparse-to-Dense Feature Matching and Visual Disambiguation
• EditJev: System One Routing for Image Editing Models
• DT-JEPA: Learning to Predict Future Representations for Decision-Making
• When to Follow the Teacher: Advantage-Based On-Policy Distillation
• UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents
• Reverse-KL Policy Distillation: Statistical Guarantees and Limits
• Rethinking Sequence Compression with Test-time Regression
• PAIR: Preference-Aligned Intent Retrieval for Premium People Recommendation at LinkedIn
• OpenAttention: One Percent Is Enough with Tail Proxies
• OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models
• MOER: Mix-Policy Optimization under Expert Prefix Guidance with Selective Rollback
• LazyFlash: Forward-Dense, Backward-Sparse Attention for Accelerated Long-Context Fine-Tuning
• Jev-DLC: Decision-Level Continual Learning from Typed Feedback
• Inside-JEPA: From Prediction to Relational Grounding in Latent World Models
• GIFT: Gradient Informed Filtering for Distillation Augmented Reasoning RL
• When Multiple Choices Are Correct: Asymmetric Probability Allocation in Jev and Other Decision Models
• Right On Q: Recovering Reasoning under Quantization via Self-Generated Trajectories
• Progressive Learning from Content to Relevance to Personalization for CTR Prediction
• PLE: Physical-Engineering's Last Exam
• MoEZip: Routing-Aware KV Cache Compression for Sparse Mixture-of-Experts LLMs
• Edit, Don't Imitate: Minimal-Edit On-Policy Distillation for Multimodal Reasoning
• CORD: Conflict-Aware Optimization for Joint RL and Distillation in Agents
• Building Auditable AI Text Detectors with Sparse Autoencoders
• Asymmetric On-Policy Distillation Learns In-Context Exploration
• Agentic World Action Models
• ALPS: Visual State Encoders for World Models
• Stratified Sparse Attention: Reducing the Sparse-to-Dense Gap at Extreme Sparsity
• SPAR: Stepwise Personalized Autoregressive Recommendation with Prefix-Balanced Semantic IDs
• RACE: RoPE-Aligned Attention Compression Based on Query-Key Activation Energies
• Projected Distillation: On Learning to Solve Hard Problems from Close, Correct Targets
• On-Policy Transition Distillation
• Object-Oriented Harness: Evolving Agent Harnesses through Object-Oriented Design
• Multimodal Reranking Is a System One Decision
• LocaLeJEPA: Maximum Entropy Representations under Geometric Priors
• Evolving Agent Interfaces
• ECHO: Efficient Cascade with Homogeneous Operators for Distillation-Guided 100K-Sequence Modeling
• Cache Engineering for Agents
• CSD: Common-Support Distillation for Cross-Tokenizer Knowledge Transfer
• Baseline and Incremental: User, Item, and Request Debiasing in Instagram Feed Ranking
• Ask to Distill: Learning to Synthesize Data for On-Policy Distillation
• Adaptive Pullback: Reweighting Supervision at Penalty Jumps in On-Policy Distillation
• Remembering Has a Price: Learning Memory Worth under Resource Constraints
• Reinforcement Learning with Visually Attributable On-Policy Self-Distillation
• MAPS: A Multi-Agent Personalized Script-Driven Video Advertising System Enhanced by Reinforcement Learning
• Kronecker Delta Rule for Long Context Linear Attention
• FLINT: Fixed-budget Late Interaction
• Don't Rewrite the SID Map: Topology Preservation for Generative Retrieval in Dynamic E-Commerce Search
• DeepSEA: A Multilingual Evaluation Benchmark for Southeast Asia
• Correctness-Dependent Asymmetric Target Distributions for Self-Distillation Policy Optimization
• T-MUON: Tensorizing Muon via t-SVD for Large Language Model Pre-training
• MemPrice: Online Memory Allocation via Adaptive Pricing for Long-Horizon Agents
• Beyond the Teacher: Understanding and Preserving Distillation Overshoot in Agentic On-Policy Distillation
• Muse Glimmer 30B: An On-Device Agentic Model
• Learning to Program Retrievals: Composing Search Primitives via Reinforcement Learning
• Aptamers as a sandbox for lab-in-the-loop agentic molecular discovery
• World of SWECraft: Agent Collaboration at Scale via Simulated Software Worlds
• Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning
• Sink-Sparse Hybrid Attention for Efficient Inference in T2I/T2V Diffusion Transformers
• SimOPD: Relational Knowledge Transfer for On-Policy Distillation
• SCOPE: Selective Certification of Program Edits for Embodied Heuristic Learning
• Preserving Reasoning Diversity in Reinforcement Learning with Verifiable Rewards via Adaptive Replay
• Persistent Planar Memory for Video World Models
• MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation
• MGE-PO: Entropy-Aware Credit Assignment for Mathematical Reasoning
• Jev-as-a-Judge for Brazilian Portuguese Information Retrieval: Relevance Assessment and Reranking
• GR-OPD: Group-Relative Sequence Credit for On-Policy Distillation
• BMuon: Budgeting Muon's Spectral Geometry for Structured Compressibility
• When the KV Cache Is Slower Than the GPU: Scaling Remote KV Tiers for Context-Parallel Prefill
• When Does Muon Actually Win? A Paired Data-, Time-, and Tuning-Budget Benchmark
• Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation
• Training-Free On-Policy Distillation
• Training Reasoning Models to Generate Diverse Scientific Ideas
• ToolBottleneckBench: Measuring Tool-Agent Performance under Controlled Pressures and Fixed Harness Budgets
• Terminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine
• TRACER: Trajectory-Aware Tool Retrieval for LLM Agents
• SoftCache: Adaptive Step Size Caching for Diffusion Transformers
• Shortcut Sensitivity in Agent-Safety Evaluations
• SciPIvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents
• SSR: Structural Signal Retriever for Scalable Agent Discovery in LLM-Orchestrated Multi-Agent Systems
• RoleQ: Role-Aligned Scales for Converting Language Models to Ternary
• Retrofitting Recurrent Vision-Language Models from Pretrained Backbones
• Rethinking Routing Exploration in Modern Mixture-of-Experts
• Repairing LLM-Generated Specifications with Deterministic Diagnosis
• RSIFlow: Adaptive Self-Improvement Workflows with an Evolving Meta-Agent
• Progress Reward Model for Reinforcement Learning via Large Language Models
• Prism-1: Position- and Rival-Independent Scoring for System One Models
• Patch where the text is new: a compression view of where byte-level models should spend compute
• PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents
• OpCredit: Operation-Aware Policy Learning for Agentic Knowledge Graph Question Answering
• OPRD: On-Policy Reminder Distillation for Long-Horizon Coding Agents
• Multi-Reward Hacking: From Aggregation-Induced Drift to Trust-Anchored Defense
• MemoProAgent: Toward Memory-Driven Proactive Assistance
• MemScope: Evolving Agent Memory Architectures through Localized Code Repair
• MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
• Librarian Agents: Training LLMs to Curate Long-Context Workspaces that Anticipate Future Queries
• Learning to Reason with Selective Internal Feedback
• Learning from Unverified Peers via Shared Residual Distillation
• LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents
• LCT: Losslessly Compressed Tensors with GPU friendly execution for Training
• LAMP: Layerwise Adaptive Mixing of Predictors for Training-Free Few-Shot Vision-Language Adaptation
• KnowGPO: Knowledge-Grouped Policy Optimization for Search Agents
• JevAgentHarness: Benchmarking, Deploying and Training Decision Models That Share Control with LLMs
• Hy-Memory-30B-A3B: Attributed Reinforcement Learning for Long-Term Memory Writing
• Harness-Delta Attribution: Dissecting Gains From Harness Evolution
• Gen4P: Video Generators are 4D Perceivers
• GACC-Bench: Benchmarking Goal-Adaptive Context Compression Ability across Large Language Models
• Fractional Power-of-Two Quantization for Efficient and Effective Multiplier-Free LLM Inference
• EliteReplay: Preserving Diverse Solutions in RL Post-Training via Quality-Diversity Search
• EgoP²C: Internalizing Plan-Perceive-Compose for Spatial Reasoning in Egocentric Video
• Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL
• DropKV: Decoupling Residual-Output Perturbation for Near-Optimal KV-Cache Eviction
• Does On-Policy Distillation Need to Complete Every Rollout?
• Diviner: Agentic RCA with Human-Governed Adaptation for Large-Scale Infrastructure
• Disagreement-Guided Weak-to-Strong Routing
• Designing Transformers for Continuous Diffusion Language Models
• Condition-Contrastive Extrapolation for On-Policy Distillation of Language Models
• Compression Does Not End at Selection: Post-Selection Positional Geometry for Sparse KV Caches
• Compatibility Matters: A Framework for Token-Level Teacher Selection in On-Policy Distillation
• CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF
• BitSparse: Compact Activation Storage for ReLU Feed-Forward Networks
• AutoSFT: An Agentic System for Automated Data Research in Supervised Fine-Tuning
• Athanor: An Agent-Native System for Autonomous Post-Training Research
• Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs
• AgentPerturb: Evaluating LLM Agent Robustness under Controlled Interaction Perturbations
• AdaLoop: Thinking with the Vision Encoder
• A team of Gemini 4 Argon agents beats Minecraft
• When Pairwise Preferences Cycle: A Reliability-Anchored Preference Walk for Video GRPO
• What Coding Agents Leave Unsaid: Measuring Task-State Gaps Between Execution and Reports
• Time-JEPA: Grounding Video World Models with Physical Time Series
• The World Won't Wait: Evaluating and Teaching Agents Under a Live Clock
• Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop
• Recursive Formation and First-Class Reuse of Compositional Capabilities
• PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference
• Open-world AI: a Tale of Three Requirements
• One-Step Prediction Fidelity Is a Poor Proxy for Planning Performance in JEPA World Models
• On-Policy Residual Learning for Diffusion Language Models
• Memory-Efficient Subspace Optimization with Residual Correction: From Momentum SGD to Muon
• Maximising Transparency in Context Compaction Improves Retention of Facts and Commitments for Lightweight Coding Agents
• Information-Gain Calibrated On-Policy Distillation for Deep Search Agents
• Gibbs-Gap Policy Optimization for Off-Policy RL with Verifiable Rewards
• Failure Is the Mother of Success: Recovering Supervision with Grounding Hindsight
• Evolving Together, Failing Apart: The Skill-Policy Co-evolution Trap in Self-Improving Agents
• EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos
• ESC: Escaping 2D Diffusion Ambiguity with Geometric Constraints for Consistent Novel View Synthesis
• Divergence-Routed On-Policy Distillation for Language Models
• Dataset-Level Regularization for Joint-Embedding Self-Supervised Learning
• CASI: Credit Assignment via Set Intervention for Agentic Retrieval
• AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks
• ARIA: Runtime Inspection and Amendment for Multi-Agent Systems
• LoOPSD: On-Policy Self-Distillation with Looped Teachers
• CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline
• WeGR: Unified Generative Retrieval for Industrial Search Systems in WeChat
• UniSGR-V2: Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation
• RSIPO: Recursive Self-Improving Policy Optimization for Agentic Reinforcement Learning via Skill-Policy Co-Evolution
• The Company a Question Keeps: Auditing Open Jev-Style Models for Retrieval
• LLM-based User-Product Relevance Evaluation for Dynamic Product Advertising
• Harmony Harness: Multi-Objective Self-Evolution of Agent Harnesses
• When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows
• Trajectory-Aware Policy Optimization: Closing the State Distribution Gap in Long-Horizon LLM RL
• Test-Time Memory Construction
• State-OPD: State-Grounded On-Policy Distillation for Multi-Turn Agents
• SmolWAM: A Latent World-Action Model for Robotic Manipulation on Edge Devices
• SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents
• ReplaySSM: Cache SSM Inputs, Not State
• RSI-Anything: Unlocking Recursive Self-Improvement with World Models
• RATED: Reliability-Aware AI Agent Tool Retrieval via Execution Dynamics
• Orthros: Phase-Aware Heterogeneous Attention for Efficient Transformers
• On the Design Space of Continuous Diffusion Language Models
• Not All Dimensions Are Equal: Outlier-Aware Attention Linearization with Grouped Quadratic Kernels
• NEAR: Neighborhood-Aware Residuals for Memory Retrieval
• Multi-Stream Attention Residuals
• Learning More by Training Less: Uncertainty-Guided Progressive Contraction for Efficient Reasoning Post-Training
• Janus: Enhancing Long-Horizon Autonomous Research with Co-Evolving World Models
• Inject a Turn, Not a Token: Learning Where to Redirect LLM Reasoning for Self-Correction
• Harness²: Recursive Agent Harnessing for an Open World
• Generalized Proximal Policy Optimization: Unifying LLM RL Algorithms from the Perspective of Divergence
• FORTE: A Production Framework for Feature Evaluation and Deprecation in Industrial Recommender Systems
• Evo2: Evolving the Evolver through Weight-Harness Co-Evolution
• EV-JEPA: Elastic V-JEPA for Any Device and Cross-Device Inference
• DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs
• Does System One Transfer? Calibrated Typed-Decision Models, Label Budgets, and Translation for Indonesian Regional Languages
• Continually Post-training Small Models via Synthetic Self-distillation
• CV-OPD: Contrastive Vision On-policy Distillation
• CAST: Credit Alignment for Cross-Tokenizer On-Policy Distillation
• Branching On-Policy Distillation
• BoundaryEvo: Data-Model Co-Evolution at the Capability Boundary for Long-Chain Reasoning Distillation
• Beyond the Final Embedding: Retrieving Memories from Token-Level Evidence
• Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation
• A Step Harder, Not a Step Too Far: The Benefits and Limits of Curriculum Difficulty Progression in GRPO
• Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals
• Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model–Skill Co-Evolution
• When Should Agents Think? Adaptive Reasoning via Cross-Turn Estimation
• When KL Regularization Misfires in Group Policy Optimization
• When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often Better
• WOVEN: Weaving Visual World Modeling into Multimodal LLMs
• VFold: Symmetry-Aware Cross-Layer Value Cache Compression
• V-CoLA: Vision Token Compression with Linear Attention
• Use and Disuse: Intent-Structured Experience Consolidation for Memory and Learning in LLM Agents
• Unsupervised Multimodal Reward Modeling
• Specialized Decision Models vs. General-Purpose LLMs: Benchmarking Jev Across Knowledge, Reasoning, and Multilingual Tasks
• SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference
• Smoothing the Top-k Exposure Boundary for Sparse Mixture-of-Experts
• Rounding in Preconditioner Space: Redesigning 4-bit AdamW Optimizer-State Quantization
• Residual Advantage: Student-Relative Teacher Guidance for RL with Verifiable Rewards
• Recursive Self-Improvement through Multi-Agent Self-Supervision
• ReCal: Calibrating Structured Pruning for On-Policy Distillation Recovery
• PIVOT: Perplexity-Informed KD-to-RL Transition Scheduling for Vertical-Domain Few-Shot Distillation
• One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails
• OnTrack: Real-Time Monitoring and Intervention in LLM Agent Trajectories via Streaming Structure-Aware Optimal Transport
• Mid-Training Language Models on Raw Video
• MetaOPD: Meta-Learned Token Weighting for On-Policy Distillation
• MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface
• Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks
• MagicBeanStalk: Escaping the RLVR Zero Advantage Trap with Recursive Data Synthesis
• Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning
• LeWAM: End-to-End World and Action Modeling with JEPAs
• GRPODropout: Less is More for Online Reinforcement Learning Rollouts
• Few-Step Generation via Data-Space Iteration
• EvoAlloc: A Self-Evolving Resource Allocation Agent for Efficient Program Evolution
• DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation
• CoTS: Competitive Token Selection for Latent Reasoning in Large Language Models
• SpecGuard: Proving a Task Is Broken Before the Agent Cheats
• EmbeddingGemma 2: Unified and Efficient Multimodal Representations
• Video Encoders Built on Image Representations
• Sentry: Learning to Recover from LLM Agent Failures at Test Time
• SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting
• LESSER: Post-Training Data Selection with Output-Layer Gradients
• FrugalEvo: Towards Cost-Aware LLM-Guided Program Evolution
• Follow the Winners: Conservative Policy Improvement with the Cross-Entropy Method for Critic-Free RFT
• Credit Where It Matters: Dependency-Aware Policy Optimization for Terminal Agents
• AdaStep: Adaptive Step Credit Weighting for Agentic Reinforcement Learning
• Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes
• VeriHarness: Scaling Agentic Verification for Long-Horizon Tasks
• VETO: Video Efficient Token Optimization for Vision Language Models
• Token-Level Video Reinforcement Learning
• TESS: Scalable, Transferable Meta-network for Data Selection Requires a Different Loss (and Why the Obvious Choice is Problematic)
• TACO: Ternary Absolute-max Column-wise One-sparse Optimizer for LLM Fine-Tuning
• Sharpening Tax in Post-Training
• Safety Must Survive Self-Improvement: Why Failures Persist and How Agents Recover
• Role-aware Heuristic Episodic Attention for Conversational LLMs
• Range-GRPO: Policy Optimization via Pairwise Relations among Reward Intervals
• RPTune: Learned Context Curation for LLM Catalog Search
• RLCPR: Rethinking Probability-Based Reinforcement Learning From Posterior Concentration
• PACE: Provenance-Aware Capability Enforcement for Tool-Using LLM Agents
• Optimizing Effective Training Time for Large-Scale Recommendation Systems
• Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation
• Not All Is Lost: Repairing Lossy User Preference States of Personalization Encoders
• Neither Black nor White: Balancing Semantic and Collaborative Signals with Graph-Informed Semantic IDs (GrIS)
• My FAULT: Self-Diagnosis as Credit Assignment in Self-Evolving Agentic Reinforcement Learning
• Mingbird: A Local-First Agent Harness Enabling Small Open Models to Complete Real Tasks
• Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents
• MWOP: Modality-aware Width-wise Operation Pruning for Efficient MLLMs
• LLM2Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them
• It Takes Workflows to Evolve Better Workflows
• IrekoGPT: Turning Structured Pruning into Post-Hoc Slimmable LLMs
• GAW-PO: Preference Optimization with Gradient-Aligned Token Weights
• Dependency-Aware Reward Shaping for Agentic Reinforcement Learning
• DeFA: Dependency-Guided Failure Attribution for LLM Agents
• Code Owns the Simulation, Jev Owns the Evaluation
• Causal Memory Policy: Making Memory Utility Identifiable by Intervening on Retrieval
• CARM: Cancellation-Aware Response Masking for LLM Reinforcement Learning
• Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
• AutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents
• Asynchronous LLM Post-Training: Group-Mass Capping and Convergence Analysis
• AgentWebRec: Compact Evidence Fusion over the Agent Web for Personalized Recommendation
• ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization
• AF-Muon: An AdamW-Free Muon Optimizer for Tied-Embedding Models
• T2SPO: Trajectory-to-Step Policy Optimization for Agentic Reinforcement Learning
• Self-Evolving Coding Rules for AI Coding Agents
• SHARPO: Segment-Level Credit Assignment for Agentic Reinforcement Learning
• MemFit: Efficient Long-Term Agentic Memory
• JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces
• HAWK: Rethinking Multimodal Drafting for Speculative Decoding
• Generative End-to-end Ad Retrieval at Douyin
• Video-RSI: Recursive Self-Improvement of Video Understanding Agents via Harness Evolution
• V-JEPA Policy: Building Effective World-Action Models on Predictive Visual Latents
• UserProxyBench: Evaluating LLM User Simulators for Agent Benchmarks and Training
• UpliftMem: Learning Set-Level Uplift for Agent Memory Retrieval
• Unlocking the Critic: Reward-Free Policy Optimization for LLM Post-Training
• Triadic Linear Attention: Three-Dimensional Recurrent States for Long-Context Sequence Modeling
• Time-Anchored Diffusion Language Models: Latent-Space Caching for Fast Generation
• Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning
• The Weakest Link: Distilling LLM Reasoning with Worst-Case Constrained Reinforcement Learning
• The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation
• STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization
• SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving
• SIPO: Unifying Reinforcement Learning with On-Policy Self-Distillation
• ReMem: Rethinking Perception and Memory in Long-Context Recommendation Agents
• Pretraining Latent Information Feedback Transformers with Teacher Supervision
• PR-OPD: Privileged Representation On-policy Self-Distillation for Agentic Reinforcement Learning
• Overcoming Scaling Limits in On-Policy Self-Distillation for LLM Reasoning
• On Trajectory-Aware Training for Masked Diffusion Language Models
• Mixture of Self-Improving Branches for Agent Harness Optimization
• Learning Meta-Skills for Agent Harness Design in Test-Time AI4AI
• Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR
• LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization
• LEGO-OPD: Factorized Teacher Composition for Multimodal On-Policy Distillation
• Interpolated Policy Distillation: A Controllable Continuum Between Off-Policy and On-Policy Distillation
• HELIX: Purified and Unified - Rethinking Feature Interaction and Sequence Modeling for Large-Scale Recommendation
• Guide, Then Let Go: Gap-Adaptive Teacher Scheduling for Sparse-Reward Agentic RL
• GRP v0.1 Technical Report
• From Dissonance to Orchestration: Teacher Intervention in On-Policy Distillation
• FlowMap-OPD: Rollout-Kernel Separation for On-Policy Distillation of Few-Step Flow-Map Generators
• DriftOPD: Sequence-Level Reverse-KL Distillation for One-Step VLA Policies
• Dr. OPD: Learning What to Follow for Optimal On-Policy Distillation of Large Language Models
• Cross-Entropy Guided Routing in Mixture-of-Experts Large Language Models
• Context Language Models
• Chinese-Jev: Bringing System One Model to Chinese-Language Tasks
• Challenges and Solutions for Bandits in the Wild: Warm-Started Mixture Bandits for Cross-Cohort Slate Recommendation
• Beyond Interaction Capacity: Estimator Scaling with Recursive Models for CTR Prediction
• AdviSD: Learning to Advise Frontier LLMs via Targeted Multi-Turn Self-Distillation
• Telescopic Language Models
• Shockingly Simple Self-retrospection Improves Agentic Models Without RL
• SAGE: Symbolic Action-Gating and Editing for LLM Task Planners
• ROSS: Relearning from Self-Generated Rollouts through Selective Supervision
• Mnemon: Raw Records, Fast Judgments, Slow Thoughts
• Measuring and Mitigating Identity-Cue Preference Drift in LLM-based Recommender Systems
• MS-GLA: Multi-Scale Gated Linear Attention for Addressing Representational Bottlenecks via Multi-Temporal Resolution
• MAS-OPD: On-Policy Distillation for Multi-Agent Systems
• Learning from Teacher Continuations at Student States
• Harness Learning Enables Generalizable Test-Time Adaptation
• GraphHCA: Closed-Form Hindsight Credit Assignment for Long-Horizon LLM Agents
• Fractional State Space Transition for Long Sequence Modeling
• EvoSkillRec: Skill-Genome Evolution for Recommender Architecture Discovery
• Dr.Credit: Rubric-Grounded Process Credit Assignment for Deep Research Agents
• Continuous Context Management
• Certified Selective Automation of LLM Agent Evaluation
• An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning
• Mend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation
• Words Speak Louder Than Order: A Behavioral Evaluation of Gemma 4
• Recursive Self-Improvement via On-Policy Distillation for Reasoning
• KuaFu: Compressing Long User Behavior into Understanding at Billion Scale
• X-Rec Technical Report
• Advancing Model Research in AgentX: Long-Horizon Autonomy for Industrial Recommender Systems
• OneTrans-V2: Unifying Retrieval, Pre-rank, and Fine-rank with One Transformer in Industrial Recommender
• Learned Cross-Task Relationships in Multi-Task Models
• PACT: From Credit Assignment to Critic Alignment
• Lightweight Ranking Heads: Accelerating Multi-Task Experimentation in Production Recommender Systems
• UNIQUE: A Unified Retrieval and Ranking System for Large-Scale Feed Recommendation
• MuSeR: Scalable Long-sequence Recommendation with Multi-interest Modeling
• Explainable Recommendations at Scale: LLM Rationales for YouTube Music Artist Discovery
• Verify, Don't Trust: Agentic Model Development for Video Discovery Retrieval at Scale
• dQwen3.5: Adapting Hybrid Language Models into Bidirectional Diffusion Language Models
• RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning
• On-Demand Attention: Efficient Long-Context Decoding with Learned Recall
• An Empirical Study of Harness Design for Coding Agents
• Trajectory Learnability for Offline On-Policy Distillation
• One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations
• Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning
• CERA-MoA: Co-Evolving Router and Agents for Mixture-of-Agents
• ASPIRE: Asynchronous Batch Self-Speculative Decoding
• A Zeroth-Order Paradigm for LLM Preference Alignment
• VideoMM: Adaptive Macro-Micro Inference for Efficient Video MLLMs
• Verifiable Social Reasoning for LLM Assistants
• Turn-level Multiscale Density Ratio Estimation for LLM Agents
• TIAO: Token Importance-Aware Policy Optimization for Text Summarization
• Style-Debiased DPO: Updating LLM Knowledge with Factuality-Aware Synthetic Preference Data
• StackTok: Accelerating VLMs Inference with Budget-Adaptive Visual Token Selection
• ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents
• RepoAtlas: Guiding Coding Agents via Evolving Multimodal Repository Views
• Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agents
• Persistent Recurrent Memory Between Transformer Layers Improves Language Model Generalization
• OPEN-1B: A Fully Auditable Training Run
• OPD-Aha: From Linguistic Momentum to Visual Reflection in Multimodal On-Policy Distillation
• LoopSpec: Pipelined Self-Speculative Decoding for Looped Transformers
• Interactive Memory Learning for Long-Term Conversations
• How Calibration Content Shapes Attention-Based Reranking
• GrowMTP: Can RL Grow Its Own Draft Head?
• ECHO: Early-layer Collaborative Hierarchical Orchestration with Bonus Logits in Speculative Decoding
• Beyond Token-Local Imitation: Reward-Compatible Temporal Credit Assignment for On-Policy Distillation
• Register Tokens for Bounded-State Reasoning in Diffusion Language Models
• LazFormer: Scaling Transformers for Industrial Recommendation via Transferable Generative Pre-training
• Generate to Explore, Select to Exploit: Aligning LLM-based Headline Generation with Personalized Recommendation
• AgentKV: Phase-Aware KV Eviction for Agentic LLMs
• Know When to Stop, Where to Restart: Accelerating Multi-Turn Agentic On-Policy Distillation
• MaD-RL: Matching Distributions for Calibrating LLMs with Reinforcement Learning
• HarnessBandit: Joint Learnability-Transferability Scheduling for Multi-Harness Agentic Reinforcement Learning
• Data-free On-policy Distillation
• SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking
• MIMA: Multi-Interest Recommendation via Multi-Positive Exclusive Assignment
• ChronicleRec: Pre-training Temporally Anchored Tokens for Lifelong User Modeling
• Beyond Vector Similarity: Hierarchical Context-Aware Graph RAG vs Standard RAG in Enterprise Code Migration
• UniRec: Cross-Stage Multi-Task Fusion with Preference Optimization for Cascaded Recommender Systems
• SenseNova-U1.5: Towards Native Unified Visual Intelligence
• SWRouter: Similarity-Contractive Window Routing for Multi-Turn Large Language Model Conversations
• SIRF: A Spec-Internalized Risk Foundation Model for Industrial Content Risk Control
• OmniKVQuant: KV Cache Quantization for Omni-LLMs
• Negative Self-Distillation: Learning to Reason by Avoiding Flaws
• Musec: MomentUm SpEctral Clipping for Stable Muon-type Training
• MAPLE: Memory-Augmented Planning with Language and Evolution
• LOCUS: Task-Aware Low-Rank Post-Training for Token-Efficient Language Generation
• Ecdysis: Efficient and Effective Training of Runtime Harnesses for LLM Agents
• Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated Data
• Caption-once, Frames-on-Demand: Visual-Need Routing for Budget-Aware Agentic Long Video Understanding
• COBRA-Skills: Contextual Bandit-Guided Evolution for Agent Skill Optimization
• Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning
• A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients
• When Synthetic Data Hurts: On Catastrophic Forgetting in Skill Retrieval for LLM Agents
• T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks
• SearchAtlas: Analyzing Agentic Search Strategies via Evidential Query Graphs
• Grounding Agent Memory: Environment-Probing Curation for Enterprise Agents
• CompassOPD: Cross-Family On-Policy Distillation via Within-Family Likelihood Shifts
• SequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching
• Procedural Graphs: Self-Evolving Execution Structures for LLM Agents
• MemForest: Efficient Agent Memory Management via EventTree Partitioning and Progressive Merging
• Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks
• Entropy-Regularized Rank-Masked Policy Optimization for Test-Time Reinforcement Learning in Code Generation
• Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation
• Distillation as Probability Transport: Routed On-Policy Distillation
• BAFF: Bid-Aware Filter Family for Mitigating Training Data Interference in RTB A/B Tests
• KVMem: Virtualizing Million-Token Agent Workspaces on a Consumer GPU
• CoSkill: Joint Reinforcement Learning of Reasoning and Meta-Skill Agents for Hierarchical Skill Evolution
• Beyond Co-purchase Relation: Evolution of Complementary Recommendations at Allegro
• BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference
• AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems
• AtomRec: Evolving Atomic Memory for Agentic Recommendation
• What Does Multi-Harness RL Learn? Credit Assignment and Portability in Coding Agents
• SiLR: Structure-Preserving Admission and Process Reward for LLM Tool Agents
• Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
• Lngram v2: Latent N-Gram Memory with Interpretable Discrete Representations
• Extremely Sparse Supervision Incentivizes Reasoning Ability
• CORAL: An LLM-Native Harness for Production Recommender Systems
• TGR: Advancing Industrial Recommendation from Generative-Paradigm Ranking toward Unified Generation and Reasoning
• From Visual Cues to Spoken Narration: Rethinking Audio Description
• From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs
• SetMIR: Multi-Interest Retrieval as Set Prediction
• CAMIE: Co-Engagement-Aware Multimodal Item Embeddings for Snap Dynamic Product Ads Retrieval
• Imagine Before Acting: Prospective Skill Identification for LLM Agents
• When Memory Takes Gradients: Collaborative Vector Memory for Agentic Recommender Systems
• What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents
• Video-OPSD: Exploiting Privileged Visual Evidence for On-Policy Self-Distillation in Video Large Language Models
• Verify Smarter, Evolve Further: Efficient Harness Evolution through Behavior-Aware Verification
• TwinKV: A Composable Repair Pass for KV Cache Eviction via Pairwise Key Redundancy
• TTPO: Test-Time Policy Optimization
• Scaling Graph Neural Networks for Friend Recommendation: Multi-Hash User Embeddings and Temporal Neighbor Sampling
• SWE-Prime: Fewer Trajectories, Better Performance
• SPT: Skills as Pre-Training Data for Agentic Language Models
• SPEAR: Distilling Domain-Adaptive Reasoning Skeletons via Sequential Symbolic Alignment in Reinforcement Learning
• RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution
• Preserving General Capabilities during Domain Specialization with Uncertainty-Calibrated MOPD
• PACE: A Unified Condense-and-Extract Paradigm for Fast VLM Inference
• Disentangling Optimization Scale from Preference Scale in DPO
• DeepRepro: State-Aware Subplanning for Paper-to-Code Reproduction in Evolving Repositories
• CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes
• Consolidating RLVR Capabilities Across Domains: A Deep Dive into Fusion Paradigms
• Boosting LLM Exploration via Weak-Model Guidance in RLVR
• Where to Look Matters: On-Policy Self-Distillation for Long-Video Understanding
• VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning
• V-Rubrics: Visual Faithfulness via Rubric-Based Reinforcement Learning
• TransRetrieval: Scaling Up Transformer-Based Retrieval for Industrial Recommendation
• TraceML: An Empirical Analysis of Human-Agent Planning in Machine Learning Development
• TOPAS: Workflow-Aware Prefix-State Scheduling for Multi-Agent LLM Serving
• SKILL.state: Scalable Long-Horizon Agent Skills
• ProgRouter: Online Progress-Guided Orchestration for Multi-Agent LLM Workflows under Quality-Cost Tradeoffs
• MLLMCLIP: Feature-Level Distillation of MLLM for Robust Vision-Language Representations
• JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution
• GRIP: Granular Reward-Guided Parameter Interpolation for Efficient Reasoning
• From Memorization to Absorption: Mixed-Policy RL for Continual Knowledge Injection
• DCEO: Direct Causal Effect Optimization for Long-Term User Value Modeling in E-commerce Search
• CaSKG: Counterfactual-Causal Skill Graphs for Scalable Agent Skill Retrieval
• AdaVDR: Adaptive Tool Use and Reflection for Video Deep Research
• WeMM-Embedding: WeChat Multi-Modal Embedding Technical Report
• TAGR: Temporally Adaptive Generative Recommendation for Industrial Live-Streaming Advertising
• SkillForge: Automated Skill Discovery and Refinement for Tool-Using Agents
• SPO++: Stabilizing Asynchronous Agentic Reinforcement Learning via Measure-Theoretic Token Correction
• SMITH: Self-Improving Tool-Using Agents through Multi-Aspect Verification
• OPDVR: On-Policy Distillation with Verifiable Rewards
• OPDSearch+: Search-Enhanced On-Policy Distillation with Reinforcement Learning
• AHEAD: Agentic Hints for Effective Agent Development
• SRPO: Self-Reflective Policy Optimization for Long-Horizon Reasoning
• Beyond the Stability-Exploration Dilemma: Environmental Regularization for LLM Policy Optimization
• AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces
• Agent-G²: Gaussian Guidance for Agentic Reinforcement Learning
• RARE: Decoupling Representation Steering from Expert Routing in Mixture-of-Experts Language Models
• From a Static Multi-Level Small Semantic Codebook to a Dynamic Single-Level Large Semantic Codebook for Generative Recommendation
• AUSO: Action-Level Unified Skill Optimization from Internalization to Utilization
• SAPO: Single-Rollout Autoregressive Policy Optimization for Agentic Reinforcement Learning
• SPADE: Self-Play in Adaptive Synthetic Executable Environments
• RTPO: Reverse-Turn Policy Optimization for Stabilizing Agentic RL Training
• OneModel: A Unified Foundation for Platform-Scale Multi-Scenario Ranking
• Beyond Imitation: Filtering On-Policy Distillation by Reasoning Progress
• PlanPO: Group Planning-Aware Policy Optimization for Multi-Turn Agentic LLMs
• Once Generated, Ranked: End-to-End Generative Slate Recommendation with Unified Semantic-Collaborative IDs
• Multimedia Asset Personalization via Multimodal Embeddings at Netflix
• TRCA: Transition-wise Rubric Credit Assignment for Long-horizon LLM Agents
• Learn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization
• Grounding AI Agents in Contracts: An Empirical Evaluation of Spec-Driven Test Generation
• HyMem: Hierarchical Context Management for Long-Horizon Agents via Information Isolation
• Fast A/B/n Testing: Exact Multi-Policy Comparison via Tree-Coupled Feedback Sharing
• DrEM: Dual-Side Robust Ensemble Ranking from Noisy User Preference Predictions in Video Recommendation
• Rubric Dropout: A Simple Way to Mitigate Reward Hacking in Rubric-as-Reward RL
• Preference Tree Optimization: Enhancing Goal-Oriented Dialogue with Look-Ahead Simulations
• LoongReflect: Boosting Long-Horizon Reflection in Search Agents via Global Perspective Distillation
• Generation as Auxiliary Supervision: Enhancing Visual Understanding at Zero Inference Overhead via Decoupled Embedding Prediction
• GCPO: Diagnosing and Constraining Subspace Geometry in Rollout RL for LLMs
• Context Blindness in DPO: Mitigating Object Hallucination in MLLMs via Context-Calibrated Preference Optimization
• Sona Technical Report
• Efficient Reinforcement Learning for Long-Horizon Tool-Use Agentic Tasks
• SR-OPSD: Self-Referenced On-Policy Self-Distillation
• OpenLoopEvolve: A Verifiable Self-Evolution Framework for Loop Policies in Long-Horizon Complex Tasks
• MetaStrategy: Generative Ranking with Executable LLM Strategies
• IntHQ: Task-Interactive Hierarchical Query on Dual-Stream Representations for Generative Recommendation
• GenRec: An LLM-Backed Recommendation Ranker at Netflix
• From Prediction to Incrementality: Causal Optimization for Large-Scale Targeting and Recommendation
• DREAM: A Dual-Loop Recommendation Evolution Framework Powered by Large Language Models
• Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension
• ConnectionMind: A General Social-Personalized Recommendation System with LLM Reasoning
• DistillCache: KL-Guided Adaptive KV-Cache Eviction for Memory-Efficient LLM Inference
• PushDualGen: Enabling LLMs to Generate Semantic IDs with Interpretable Copy for Industrial Push Recommendation
• Teacher Retains Full Tokens, Student Merges Efficiently: TM20K for E-Commerce Sequence Modeling in Ad Recommendation
• ResidencyRL: Reinforcement Learning in Simulated Clinical Environments
• Coupling Planning with Episodic Memory in LLM Agents for Software Issue Resolution
• Autonomy-of-Heads: Data-Free Sparse Attention from Frozen Query-Key Geometry
• When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents
• U-OPSD
• RRC: Unlocking Generative Reward Models in LLM Reinforcement Learning via Ranking-Based Reward Construction
• RP-OPSD
• On-Policy Delta Distillation for Multilingual Math Reasoning
• MACRO: Markov Chain Routing of Transformer Layers
• Learning Globally Reusable Skills for Coding Agents
• Hierarchical Latent Prediction for Language Models
• HarnessOpt-Bench: Evaluating LLMs at Harness Optimization
• Gryphon-v2: One Model in Place of a Cascade — Generate-and-Rank Recommender with Rollout Distillation
• EnvACE
• Detecting Safety Training Modification in Language Models via Activation Analysis
• DASH
• Contextual Information Policy Optimization for Search Agents
• CodeGrep: An RL-Trained Retrieval Agent for LLM Coding Agents
• BaKron: Efficient Quantization with Kronecker-Factored Hessians
• AgentOPSD
• Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes
• State2State: Environment-Derived Mid-Training for LLM Agents
• SpecRoll: Fast-Slow Verifier-Feedback Adaptation for Speculative Reinforcement Learning Rollouts
• QEvict: Recoverable Quantized KV Eviction for Attention-Drift-Robust Long-Context Decoding
• Optimizing What Policies Learn From: Recoverability-Aware Rollout Intervention Learning
• OCSD
• Multi-Objective Ranking for Live-Streaming: Balancing Fresh and Delayed Signals with Segment-Aware Targeting
• MemoryCPT: An End-to-End Agent Memory Framework for Cost-Performance Trade-off
• EvoHarness-RL: Learning Self-Evolving Runtime Harness for Long-Horizon LLM Agents
• DEGR: Dual Exploration-Driven Generative Re-Ranking for Adaptive Cross-Request Context Bridging
• DBLast: Dependent Block Drafting for Stochastic Speculative Decoding
• VerMem
• ToolLIFT: Lifting Tool-Specific Trajectories into Function-Level Graphs for Generalizable Tool Planning
• TaskPress: Query-Agnostic KV Cache Compression via Task-Guided Pruning
• SMOPD: Multi-Reward Reinforcement Learning via Specialize-and-Merge Online Policy Distillation
• LLM-Derived Priors for Thompson Sampling in Cold-Start Comment Recommendation
• CausalOPD: First-Wrong-Step Supervision for Distilling Causal Chain Reasoning
• ADRS
• SPEAR: Selection-aware Personalized End-to-end Adaptive Rewriting and Retrieval for Community Search
• Role-Decoupled Attention Residuals
• PCSD
• Learning What to Remember: Test-Time Training via Context Distillation
• Knowledge–Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation
• HindSearch: Trajectory-Level Hindsight Critique for Search-Augmented Reinforcement Learning
• Douyin Multimodal Embedding Model Technical Report
• DART: Decoded Attention over Recurrent States for Efficient Long-Context Sequence Modeling
• CoEvo-Mem
• Beyond On-Policy Exploration: Integrating External Policy Rollouts for Reinforcement Learning in Diffusion Language Models
• A Self-Triggered Agentic Push Recommendation System
• Hierarchical Residual Policy Optimization for Generative Recommendations
• Distill Where You Fail: Recovering Learning Signals of Negative RL-Groups from Adaptive Teacher Guidance
• TransX: Scaling Transformer-based Recommendation via Behavioral and Serving Stream Crossings
• TransMem: Transforming Hidden States into Memory for Large Language Models
• RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems
• HyperAgent: Planning and Acting over Tool-Schema Hypergraphs for Tool-Use LLM Agents
• GALA: Generative Aligned Learning for Adaptive Multimodal Representation in the Taobao Shangou Recommender System
• β-OPSD
• WIDE: Boosting Adaptive LLM Inference via Token-level Dynamic Width Pruning
• VAD
• TAPO
• ReToken: One Token to Improve Vision–Language Models for Visual Retrieval
• ROCS: Request-Oriented Compute Sharing for Efficient Large-Scale Recommendation
• OSReward / OS-Shepherd
• MANTA: Multi-Agent Network Topology Adaptation for Self-Evolving Multi-Agent Systems
• LLM-Based Generative Retrieval for Snapchat Content Recommendation
• Heterogeneous Ranking in Industrial-Scale Recommender Systems: A Case Study
• Group-Reflective Self-Distillation
• From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation
• Flux-OPD
• FinanceHarness: Autonomous Financial Deep Research Framework
• Contrastive Reinforced Policy Optimization via Privileged Self-Distillation
• CCFormer: Efficient Cross-Field Interaction and Hierarchical Sequence Compression for Industrial Recommendation at Tencent
• Building a User Foundation Model for the Open Web
• SkillRise
• SERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement Learning
• ReCo
• PSG: Pair-Space Generation for Efficient Generative Reranking
• OneShot: Index-in-Ranking with Neural Scoring for Large-Scale Retrieval
• DIRECTOR: Dynamic Index-based Recommendation with Transport-Optimized Retrieval
• ClockRoPE: Random Fourier Rotations for Temporal Routine Modeling
• CAM-DF
• ASARL: Autonomous Social-Aware Relevance Learning for QQ Search
• UniMem
• TWICE: Two-Clock, Two-Window Learning for Long-Horizon Conversion Prediction in Online Advertising
• Reward Guided Decoding for Generative Recommendation
• Relay-OPD
• Reinforcement Learning for Code Optimization
• RecoReward: Recommender-Guided Multimodal Description Generation for Recommendation
• Penelope: Localized Latent Recurrence for Efficient Structured Reasoning
• HiSkill
• CoRT
• CAST
• Beyond Single-Episode Optimization: Sliding-Window Aware Generative Auto-Bidding for Long-Term Advertising Effectiveness
• What do Reward Models Memorize?
• Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence
• SpecFormer: Mitigating Embedding and Attention Collapse via Spectral-Aware Transformer for Recommendation
• OxygenREC-v2: Internalizing Discrimination into Generative Recommendation
• Mosaic: A Fleet of User Embedding Specialists for Recommendation at Meta
• Memory Layer: Train the In-Model Cache for Recommendation Models
• DataOrchestra: Learning to Orchestrate Per-Example Curation of Pretraining Data
• CORE: A Unified Cascaded Ordinal Relevance Estimation Framework for E-commerce Search
• Tokens are All You Need: Dual-purpose Semantic IDs for Achieving LLM-Level I/O Efficiency in Recommendation Systems
• Melo: A Production LLM-Powered Music Recommendation Agent
• Breaking the Loop: An Empirical Comparison of Strategies for Novelty and Freshness in YouTube Music
• EGR: Embedding-Native Generative Retrieval with a Shared LLM
• Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context
• PinEqualizer: Full Funnel Content Exploration and Debiasing System at Pinterest
• Parameter-free Adaptive Sparse Attention via Compression-Based Content Selection
• Naju: A Native Discrete State-Space Model with Independent Retention and Writing for Long-Sequence Memory
• Hilbert Operator for Progressive Encoding (HOPE): A Mathematical Framework for Deconstructing Learned Representations in Deep Networks
• Bridging the Structural Gap: Adapting Autoregressive Generation for Recommendation
• Anti-Periodic Positional Encoding: Möbius Boundary Conditions Make In-Context Retrieval Reliable
• Adaptive Depth Sparse Framework: Similarity-Driven Resource Allocation for Pre-Trained LLMs
• TCR
• GaugeQuant: Online Learning of Quantization-Optimal Bases from LLM Symmetries
• Co-Evolving LLM Evaluators and Policies via DynamicRubric
• TSGR: Taobao Search Generative Retrieval
• PinDCO: Whole-Page Aware Dynamic Creative Optimization at Scale
• Off-Context GRPO: Learning to Reason on Hard Problems using Privileged Information
• ISO: An RLVR-Native Optimization Stack
• AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents
• RecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender Systems
• RAMP: Robust Ad Recommendation Under Limited Personalized-Feature Availability via Masking and Alignment Pathways
• PPL-Factory: Task-Aware and Budget-Aware Data Selection from Language Modeling to Reasoning
• C²KV: Compressed and Composable KV Cache Reuse for Efficient LLM Inference
• Convolution for Large Language Models
• WHALE: A Scalable Unified Model for Recommendation with Wukong-HSTU Architecture
• Uncertainty as Remedy: Mitigating Satisfaction Label Bias in Short Video Multi-Objective Ensemble Ranking
• Pin-SCALE: Semantic Cascading and Alignment Learning for Engagement-Aware IDs in Cold-Start Recommendations
• Distilled RL
• OpenLanguageModel: Readable and Composable Small-Language-Model Pretraining for Education and Research
• RecGPT-V3 Technical Report
• RECAP: Feedback-Driven Streaming Semantic User Profiles for Short-Video Recommendation
• SEED
• Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers
• LLM-Based Re-Ranking for Real Estate Search
• TMallGS: Scaling Unified Feature and Sequence Modeling for Generative E-commerce Search
• Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents
• Long-term User Engagement Optimization through Model-agnostic Downstream Rewards Learning
• Long-History User Transformers for Real-Time Ad Ranking
• Adaptive Ad Load Design for Sponsored Search Markets: Evidence, Theory, and Deployment
• SlimPer: Make Personalization Model Slim and Smart
• Proximity Features: Privacy-Compliant Cold-Start Personalization at Airbnb
• Not Only NTP: Extending Training Signal Coverage for Generative Recommendation
• MESH: Scaling Up Retrieval with Heterogeneous Content Unification
• Learning to Forget: Satiation-Aware Long-Sequence Transducers for Mitigating Post-Purchase Redundancy
• Deep-learning Causal Retrieval Optimization for Efficient e-commerce Distribution in Pinterest
• Cheaper is Better: A Discount-Aware Network for Conversion Rate Prediction in E-commerce Recommendation System
• Guess Where You Go: Generative Next Point-of-Interest Recommendation in Amap
• ZoRRO: A Zero-Weight Personalized Recommender System for Scalable News Recommendation
• RIPO
• Multilingual Semantic Retrieval for Apple Music Search
• ARMOR
• TACO
• TurnOPD
• Turning Off-Policy Tokens On-Policy: A Plug-in Approach for Improving LLM Alignment
• Prompt Generation Technical Report
• GenPage: Towards End-to-End Generative Homepage Construction at Netflix
• Editing Everything Everywhere All at Once
• POEM: Partial-Order Enhanced Real-Time Sequential Modeling for Recommendation
• MOPD
• NEXT: Reasoning-Driven Video Recommendation via a Vision-Language Model
• AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems
• UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation
• NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems
• Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale
• TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems
• Cross-Lingual Exploration for Parametric Knowledge
• CoBA-RL
• STARE
• JourneyFormer: Encoding Airbnb Guest Journey with Sequence Modeling
• RankGraph-2: Lifecycle Co-Design for Billion-Node Graph Learning in Recommendation
• PIANO: Personalized Reranking via Information Aggregation Node for Music Search Optimization
• OneRank: Unified Transformer-Native Ranking Architecture for Multi-Task Recommendation
• EvoRec: Self-Evolving Agentic Recommender Systems
• KPop
• MVEB: Massive Video Embedding Benchmark
• MiniMax Sparse Attention
• Atomic Intent Reasoning: Bringing LLM Semantics to Industrial Cross-Domain Recommendations
• ToolRec: Calibrated Preference Alignment for Query Recommendation in On-Device Assistants
• SSRLive: Live Streaming Recommendation with Dynamic Semantic ID
• OPDLM
• Taiji: Pareto Optimal Policy Optimization with Semantics-IDs Trade-off for Industrial LLM-Enhanced Recommendation
• POLARIS: Guiding Small Models to Write Long Stories
• Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories
• XLGoBench: Detecting cross-lingual skill gaps with algorithmic tasks
• Synthetic Data from Cross-Domain Events for Large-Scale Recommendation Systems
• Formalizing Mathematics at Scale
• Stochastic Primal-Dual Decoding for Multiobjective Generative Recommender Systems
• MuChator: Enabling Active Music Discovery via Conversational Music LLMs in Douyin Music
• L2Rec: Towards Dual-View Understanding of LLMs for Personalized Recommendation
• Causal Representation Learning for Generalisable Recommendation
• ScientistOne: Towards Human-Level Autonomous Research via Chain-of-Evidence
• From Item-Only to Query-Item: Query-Conditioned Generative Search with QGS in Quark
• DeGRe: Listwise Generative Reranking with Offline Lookahead Distillation
• TubiFM: Unified Item, Carousel, and Search Ranking for Streaming Discovery
• HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval
• From Head to Tail: Asymmetric Knowledge Transfer in Long-tail Recommendation with Generative Semantic IDs
• PEARL: Unbiased Percentile Estimation via Contrastive Learning for Industrial-Scale Livestream Recommendation
• Memory Grafting: Scaling Language Model Pre-training via Offline Conditional Memory
• FLUID: From Ephemeral IDs to Multimodal Semantic Codes for Industrial-Scale Livestreaming Recommendation
• Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search
• PROMPTS: Performance Optimization via Multi-Agent Planning for LLM Training and Serving
• GPRL
• DADF: A Distribution-Aware Debiasing Framework for Watch-Time Regression in Recommender Systems
• Policy-Grounded Dynamic Facet Suggestions for Job Search
• Customizing an LLM for Enterprise Software Engineering
• Efficient Generative Retrieval for E-commerce Search with Semantic Cluster IDs and Expert-Guided RL
• ConSPO
• Missing Old Logits
• Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training
• UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence
• A Production-Ready RL Framework for Personalized Utility Tuning with Pareto Sweeping in Pinterest Recommender Systems
• Superintelligent Retrieval Agent: The Next Frontier of Agentic Retrieval
• Effective Knowledge Transfer for Multi-Task Recommendation Models
• Bridging Passive and Active: Enhancing Conversation Starter Recommendation via Active Expression Modeling
• RecGPT-Mobile: On-Device Large Language Models for User Intent Understanding in Taobao Feed Recommendation
• Beyond Item IDs: Scaling Short-Form-Video Recommendation via Semantic-Native Long Sequence Modeling
• Unified Value Alignment for Generative Recommendation in Industrial Advertising
• LWGR: Lagrangian-Constrained Personalized World Knowledge for Generative Recommendation
• LLM Retrieval for Stable and Predictable Ad Recommendations
• Fine-Tuned LLM as a Complementary Predictor Improving Ads System
• A General Framework for Multimodal LLM-Based Multimedia Understanding in Large-Scale Recommendation Systems
• Optimized Deferral for Imbalanced Settings
• Mind the Gap: Structure-Aware Consistency in Preference Learning
• From Local Indices to Global Identifiers: Generative Reranking via Global Action Space
• Beyond Static Collision Handling: Adaptive Semantic ID Learning for Multimodal Recommendation at Industrial Scale
• ProEval: Proactive Failure Discovery and Efficient Performance Estimation for Generative AI Evaluation
• CS3: Efficient Online Capability Synergy for Two-Tower Recommendation
• AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization
• StepPO
• RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems
• Modular Representation Compression: Adapting LLMs for Efficient and Effective Recommendations
• FUSE: Ensembling Verifiers with Zero Labeled Data
• NaijaS2ST: A Multi-Accent Benchmark for Speech-to-Speech Translation in Low-Resource Nigerian Languages
• GroupDPO: Memory-Efficient Group-Wise Direct Preference Optimization
• GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation
• UniRec: Bridging the Expressive Gap between Generative and Discriminative Recommendation via Chain-of-Attribute
• Lightning OPD
• Efficient Retrieval Scaling with Hierarchical Indexing for Large Scale Recommendation
• SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling
• SID-Coord: Coordinating Semantic IDs for ID-based Ranking in Short-Video Search
• SEARL
• LogAct: Enabling Agentic Reliability via Shared Logs
• Dual-Rerank: Fusing Sequential Dependencies and Utility for Generative Reranking
• CROP: Token-Efficient Reasoning in Large Language Models via Regularized Prompt Optimization
• TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems
• PaperOrchestra: A Multi-Agent Framework for Automated AI Research Paper Writing
• MBGR: Multi-Business Generative Recommendation
• Reasoning-Driven Synthetic Data Generation and Evaluation
• RCLRec: Reverse Curriculum Learning for Modeling Sparse Conversions in Generative Recommendation
• Switch Attention: Towards Dynamic and Fine-grained Hybrid Transformers
• AIRA2: Overcoming Bottlenecks in AI Research Agents
• UniScale: Synergistic Entire Space Data and Model Scaling for Search Ranking
• GateSID: Adaptive Gating for Semantic-Collaborative Alignment in Cold-Start Recommendation
• TagLLM: A Fine-Grained Tag Generation Approach for Note Recommendation
• SaFRO: Satisfaction-Aware Fusion via Dual-Relative Policy Optimization for Short-Video Search
• GenFacet: End-to-End Generative Faceted Search via Multi-Task Preference Alignment in E-Commerce
• AIGQ: An End-to-End Hybrid Generative Architecture for E-commerce Query Recommendation
• Memento-Skills
• HyperAgents
• Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify
• Large Reasoning Models Struggle to Transfer Parametric Knowledge Across Scripts
• POLCA: Stochastic Generative Optimization with LLM
• Efficient, Property-Aligned Fan-Out Retrieval via RL-Compiled Diffusion
• SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders
• Not All Candidates are Created Equal: Heterogeneity-Aware Pre-ranking
• Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal Attacks
• Constraint-Aware Generative Re-ranking for Multi-Objective Optimization in Advertising Feeds
• PinCLIP: Large-scale Foundational Multimodal Representation at Pinterest
• OneRanker: Unified Generation and Ranking with One Model in Industrial Advertising Recommendation
• IDProxy: Cold-Start CTR Prediction for Ads and Recommendation at Xiaohongshu with Multimodal LLMs
• Theoretical Perspectives on Data Quality and Synergistic Effects in Pre- and Post-Training Reasoning Models
• Beyond the Flat Sequence: Hierarchical and Preference-Aware Generative Recommendations
• Zero-shot Cross-domain Knowledge Distillation: A Case study on YouTube Music
• Stop Treating Collisions Equally: Qualification-Aware Semantic ID Learning for Recommendation at Industrial Scale
• Learning to Reflect and Correct: Towards Better Decoding Trajectories for Large-Scale Generative Recommendation
• Vectorizing the Trie: Efficient Constrained Decoding for LLM-based Generative Retrieval on Accelerators
• U-Mem
• Prompt-Level Distillation: A Non-Parametric Alternative to Model Fine-Tuning for Efficient Reasoning
• HiSAC: Hierarchical Sparse Activation Compression for Ultra-long Sequence Modeling in Recommenders
• Generative Pseudo-Labeling for Pre-Ranking with LLMs
• Bending the Scaling Law Curve in Large-Scale Recommendation Systems
• A Theoretical Framework for Modular Learning of Robust Generative Models
• A Long-term Value Prediction Framework In Video Ranking
• Rethinking ANN-based Retrieval: Multifaceted Learnable Index for Large-scale Recommendation System
• Learning Personalized Agents from Human Feedback
• ConvApparel: A Benchmark Dataset and Validation Framework for User Simulators in Conversational Recommenders
• Climber-Pilot: A Non-Myopic Generative Recommendation Model Towards Better Instruction-Following
• RQ-GMM: Residual Quantized Gaussian Mixture Model for Multimodal Semantic Discretization in CTR Prediction
• RGAlign-Rec: Ranking-Guided Alignment for Latent Query Reasoning in Recommendation Systems
• Jointly Optimizing Debiased CTR and Uplift for Coupons Marketing: A Unified Causal Framework
• CAPTS: Channel-Aware, Preference-Aligned Trigger Selection for Multi-Channel Item-to-Item Retrieval
• Awakening Dormant Users: Generative Recommendation with Counterfactual Functional Role Reasoning
• Online Advertising with Spatial Interactions
• OPCD
• Compress, Cross and Scale: Multi-Level Compression Cross Networks for Efficient Scaling in Recommender Systems
• An Industrial-Scale Sequential Recommender for LinkedIn Feed Ranking
• Compute Only Once: UG-Separation for Efficient Large Recommendation Models
• CADET: Context-Conditioned Ads CTR Prediction With a Decoder-Only Transformer
• SMES: Towards Scalable Multi-Task Recommendation via Expert Sparsity
• SARM: LLM-Augmented Semantic Anchor for End-to-End Live-Streaming Ranking
• Kunlun: Establishing Scaling Laws for Massive-Scale Recommendation Systems through Unified Architecture Design
• DiffuReason: Bridging Latent Reasoning and Generative Refinement for Sequential Recommendation
• PIT: A Dynamic Personalized Item Tokenizer for End-to-End Generative Recommendation
• ML-DCN: Masked Low-Rank Deep Crossing Network Towards Scalable Ads Click-through Rate Prediction at Pinterest
• CausalArmor: Efficient Indirect Prompt Injection Guardrails via Causal Attribution
• MSN: A Memory-based Sparse Activation Scaling Framework for Large-scale Industrial Recommendation
• MDL: A Unified Multi-Distribution Learner in Large-scale Industrial Recommendation through Tokenization
• TokenMixer-Large: Scaling Up Large Ranking Models in Industrial Recommenders
• Statistical Learning from Attribution Sets
• AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents
• LUSPO
• DOS: Dual-Flow Orthogonal Semantic IDs for Recommendation in Meituan
• MemSkill
• MARS: Modular Agent with Reflective Search for Automated AI Research
• Unifying Ranking and Generation in Query Auto-Completion via Retrieval-Augmented Generation and Multi-Objective Alignment
• SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress
• S-GRec: Personalized Semantic-Aware Generative Recommendation with Asymmetric Advantage
• Generative Recommendation for Large-Scale Advertising
• Zenith: Scaling up Ranking Models for Billion-scale Livestreaming Recommendation
• OneMall: One Model, More Scenarios -- End-to-End Generative Recommender Family at Kuaishou E-Commerce
• Mobility-Embedded POIs: Learning What A Place Is and How It Is Used from Human Movement
• ECO: Quantized Training without Full-Precision Master Weights
• Towards End-to-End Alignment of User Satisfaction via Questionnaire in Video Recommendation
• PEARL
• Dynamic Cogeneration of Bug Reproduction Test in Agentic Program Repair
• S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation
• OPSD
• Unleashing the Potential of Sparse Attention on Long-term Behaviors for CTR Prediction
• LifeAgentBench: Benchmarking LLMs for Long-Horizon, Cross-Dimensional Lifestyle Health Reasoning
• Hierarchical Contextual Uplift Bandits for Catalog Personalization
• Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning
• CoReflect: A Reflective Co-Evolution Framework for Improving Conversational Evaluation
• Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models
• Applying Embedding-Based Retrieval to Airbnb Search
• PROMISE: Process Reward Models Unlock Test-Time Scaling Laws in Generative Recommendations
• Rethinking Multi-objective Ranking Ensemble in Recommender System: From Score Fusion to Rank Consistency
• Cold-Starting Podcast Ads and Promotions with Multi-Task Learning on Spotify
• mHC: Manifold-Constrained Hyper-Connections
• HiGR: Industrial-Scale Hierarchical Generative Slate Recommendation Framework in Tencent
• SAGE
• RecGPT-V2 Technical Report
• Online IcePop
• Stabilizing RL with LLMs
• Agent0
• Agent-R1
• DualGR: Generative Retrieval with Long and Short-Term Interests Modeling
• IcePop
• From Reasoning LLMs to BERT: A Two-Stage Distillation Framework for Search Relevance
• Watch and Learn: Learning to Use Computers from Online Videos
• LEGOMem
• Detecting Post-generation Edits to Watermarked LLM Outputs via Combinatorial Watermarking
• When LLMs Benchmark Themselves: Deconstructing Self-Bias in Automated Evaluation
• Judging with Confidence: Calibrating Autoraters to Preference Distributions
• IntSR: An Integrated Generative Framework for Search and Recommendation
• OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking
• Deep Reinforcement Learning for Ranking Utility Tuning in the Ad Recommender System at Pinterest
• MPFormer: Adaptive Framework for Industrial Multi-Task Personalized Sequential Retriever
• MUA-RL
• CHORD
• GPPO
• GLiClass: Generalist Lightweight Model for Sequence Classification Tasks
• ToolGrad: Efficient Tool-Use Dataset Generation with Textual Gradients
• TIS
• Agent Lightning
• SaviorRec: Semantic-Behavior Alignment for Cold-Start Recommendation
• OneRec-V2 Technical Report
• Model Directions, Not Words: Mechanistic Topic Models Using Sparse Autoencoders
• MemTool
• GSPO
• Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty
• Click A, Buy B: Rethinking Conversion Attribution in E-Commerce Recommendations
• Scaling Recommender Transformers to One Billion Parameters
• SPIRAL
• CISPO / MiniMax-M1
• Macro Graph of Experts for Billion-Scale Multi-Task Recommendation
• RADAR: Recall Augmentation through Deferred Asynchronous Retrieval
• SCAR: Shapley Credit Assignment for More Efficient RLHF
• INTUITOR
• WebAgent-R1
• GiGPO
• A Generative Re-ranking Model for List-level Multi-objective Optimization at Taobao
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
• Absolute Zero
• Towards Large-scale Generative Ranking
• RAGEN
• TTRL
• LUFFY
• ToolRL
• ReTool
• VAPO
• SmolVLM: Redefining small and efficient multimodal models
• DeepResearcher
• PinRec: Outcome-Conditioned, Multi-Token Generative Retrieval for Industry-Scale Recommendation Systems
• Dr. GRPO
• DAPO
• Search-R1
• SOPBench: Evaluating Language Agents at Following Standard Operating Procedures and Constraints
• Sparse Meets Dense: Unified Generative Recommendations with Cascaded Sparse-Dense Representations
• Muon is Scalable for LLM Training
• FilterLLM: Text-To-Distribution LLM for Billion-Scale Cold-Start Recommendation
• SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
• MoBA: Mixture of Block Attention for Long-Context LLMs
• Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention
• FuXi-α: Scaling Recommendation Model with Feature Interaction Enhanced Transformer
• LOOP
• MIM: Multi-modal Content Interest Modeling Paradigm for User Behavior Modeling
• Unlocking Scaling Law in Industrial Recommendation Systems with a Three-step Paradigm based Large User Model
• SessionRec: Next Session Prediction Paradigm For Generative Sequential Recommendation
• Bursting Filter Bubble: Enhancing Serendipity Recommendations with Aligned Large Language Models
• AdaF²M²: Comprehensive Learning and Responsive Leveraging Features in Recommendation System
• REINFORCE++
• Byte Latent Transformer: Patches Scale Better Than Tokens
• PRECISE: Pre-training Sequential Recommenders with Collaborative and Semantic Information
• Balancing Efficiency and Effectiveness: An LLM-Infused Approach for Optimized CTR Prediction
• Hymba: A Hybrid-head Architecture for Small Language Models
• LEADRE: Multi-Faceted Knowledge Enhanced LLM Empowered Display Advertisement Recommender System
• TWIN V2: Scaling Ultra-Long User Behavior Sequence Modeling for Enhanced CTR Prediction at Kuaishou
• OpenHands
• Contextual Distillation Model for Diversified Recommendation
• Counteracting Duration Bias in Video Recommendation via Counterfactual Watch Time
• SimPO
• SWE-agent
• Knowledge Adaptation from Large Language Model to Recommendation for Practical Industrial Application
• Data Mixing Laws: Optimizing Data Mixtures by Predicting Language Modeling Performance
• ORPO
• NoteLLM: A Retrievable Large Language Model for Note Recommendation
• Breaking the Length Barrier: LLM-Enhanced CTR Prediction in Long Textual User Behaviors
• A Large Language Model Enhanced Sequential Recommender for Joint Video and Comment Recommendation
• RLOO
• DeepSeekMath / GRPO
• KTO
• Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
• Self-Rewarding LM
• SPIN
• Math-Shepherd
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces
• GAIA
• Large Language Model based Long-tail Query Rewriting in Taobao Search
• IPO
• ReMax
• MemGPT
• SteerLM
• LATS
• RLAIF
• AutoGen
• MetaGPT
• GKD
• MiniLLM
• AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
• Towards Open-World Recommendation with Knowledge Augmentation from Large Language Models
• Let's Verify Step by Step
• ReWOO
• DPO
• Voyager
• ToolBench
• GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints
• CRITIC
• Tree of Thoughts
• SLiC-HF
• DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining
• Visual Instruction Tuning
• RAFT
• RRHF
• Generative Agents
• CAMEL
• Self-Refine
• HuggingGPT
• Reflexion
• ART
• Toolformer
• BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
• Constitutional AI
• Fast Inference from Transformers via Speculative Decoding
• PAL
• ReAct
• MRKL
• M6-Rec: Generative Pretrained Language Models are Open-Ended Recommender Systems
• SayCan
• InstructGPT / PPO-RLHF
• WebGPT
• Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
• RoFormer: Enhanced Transformer with Rotary Position Embedding
• Learning Transferable Visual Models From Natural Language Supervision
• Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity
• Progressive Layered Extraction (PLE): A Novel Multi-Task Learning Model for Personalized Recommendations
• DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systems
• Search-based User Interest Modeling with Lifelong Sequential Behavior Data for CTR Prediction
• Behavior Sequence Transformer for E-commerce Recommendation in Alibaba
• Deep Interest Evolution Network for Click-Through Rate Prediction
• Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts
• Entire Space Multi-Task Model: An Effective Approach for Estimating Post-Click Conversion Rate
• DeepFM: A Factorization-Machine based Neural Network for CTR Prediction
• Deep Neural Networks for YouTube Recommendations
• Wide & Deep Learning for Recommender Systems
• TransAct V2: Lifelong User Action Sequence Modeling on Pinterest Recommendation
• Structuring and Tokenizing Distributed User Interest Context for Generative Recommendation
• Self-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents
• Self-Attentive Sequential Recommendation
• Recommender Systems with Generative Retrieval
• Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models
• Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval
• RankMixer: Scaling Up Ranking Models in Industrial Recommenders
• PinFM: Foundation Model for User Activity Sequences at a Billion-scale Visual Discovery Platform
• PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations
• OneTrans: Unified Feature Interaction and Sequence Modeling with One Transformer in Industrial Recommender
• OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment
• MixFormer: Co-Scaling Up Dense and Sequence in Industrial Recommenders
• Memento: Personalized RAG-Style Long-Retention Data Scaling for Online Ads Recommendation
• LONGER: Scaling Up Long Sequence Modeling in Industrial Recommenders
• LLaTTE: Scaling Laws for Multi-Stage Sequence Modeling in Large-Scale Ads Recommendation
• HyFormer: Revisiting the Roles of Sequence Modeling and Feature Interaction in CTR Prediction
• Divergence Meets Consensus: A Multi-Source Negative Sampling Framework for Sequential Recommendation
• Deep Interest Network for Click-Through Rate Prediction
• CMSL: Constructive Multi-Sequence Learning for Recommendation Systems
• AutoformBench: A Lean Formalization Benchmark for Advanced Mathematics
• Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations