HuggingFace Papers

• HuggingFace Papers Today Week Month
• 共 105 条
• 1 Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills

  2609.02749 ▲ 529
• 2 StudentSim: Training LLM-based Student Simulators

  2609.01591 ▲ 483
• 3 Compile by Training: Turning Natural-Language Specifications into Local Neural Functions

  2609.04199 ▲ 376
• 4 Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

  2609.00111 ▲ 376
• 5 Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

  2609.04148 ▲ 281
• 6 HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?

  2609.01437 ▲ 256
• 7 LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes

  2609.03796 ▲ 228
• 8 Aspire: Can Models Self-Evolve from Vague Goals?

  2608.31111 ▲ 222
• 9 Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning

  2609.03430 ▲ 168
• 10 Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training

  2608.26730 ▲ 149
• 11 SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models

  2609.02886 ▲ 143
• 12 Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement

  2608.31046 ▲ 143
• 13 RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning

  2609.03199 ▲ 116
• 14 Lucida: Parse, Generate, and Place for Composable Real-to-Sim Scene Modeling

  2608.30821 ▲ 116
• 15 EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction

  2609.02783 ▲ 115
• 16 LatentPress: Context Compression Beyond Text and Vision

  2609.01507 ▲ 111
• 17 Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems

  2609.02750 ▲ 102
• 18 SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

  2609.01343 ▲ 99
• 19 DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution

  2608.31106 ▲ 99
• 20 Rethinking On-Policy Distillation of Large Language Models II: One Training Example

  2609.04172 ▲ 82
• 21 Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM

  2609.04098 ▲ 76
• 22 It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning

  2609.00638 ▲ 72
• 23 Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States

  2609.04196 ▲ 66
• 24 GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modeling

  2608.29335 ▲ 66
• 25 Language Models Can Control Their Own Attention

  2609.02737 ▲ 65
• 26 UI-Venus-2 Technical Report

  2609.00028 ▲ 62
• 27 On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability

  2608.30320 ▲ 53
• 28 Iris: Climbing to the Search Frontier

  2609.04304 ▲ 51
• 29 ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training

  2609.00188 ▲ 51
• 30 H3-World: Turning Language Understanding into World Control

  2609.01560 ▲ 50
• 31 Normalized Low-Rank Adaptation

  2608.31036 ▲ 50
• 32 Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction

  2609.04201 ▲ 46
• 33 Editable Visual Design

  2609.04034 ▲ 42
• 34 S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

  2608.31100 ▲ 38
• 35 PaperGym: Rubric-Centered Evolution for Research-Plan Generation

  2608.31119 ▲ 38
• 36 Motion-Omni: End-to-End Joint Speech and Full-Body Motion for Spoken Dialogue

  2609.04250 ▲ 38
• 37 Hi-Q: Hierarchical Evidence-guided Query Refinement for Multi-Hop Question Answering

  2608.30468 ▲ 34
• 38 The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation

  2609.02367 ▲ 33
• 39 From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix

  2609.01572 ▲ 33
• 40 On the Design Fundamentals of Pixel Text Representation Learning

  2609.01147 ▲ 32
• 41 LightNav-0: Eliciting VLM Spatial Intelligence for Generalist Embodied Navigation

  2608.30935 ▲ 32
• 42 WHALE: A Simple Recipe for Joint Harness-Weight Optimization

  2609.00196 ▲ 31
• 43 SHAPE of Chain-of-Thought in Math Reasoning

  2608.28600 ▲ 31
• 44 Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding

  2609.04131 ▲ 30
• 45 Last Translation Benchmark

  2609.04173 ▲ 30
• 46 Dr. Claw: An AI Scientist Workspace for Vibe Research

  2609.00365 ▲ 30
• 47 CogEvol: Towards Efficient and Reliable Learning Environment Generation

  2608.30968 ▲ 30
• 48 RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests

  2608.27831 ▲ 29
• 49 Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

  2608.21450 ▲ 29
• 50 Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence

  2608.31075 ▲ 28
• 51 Evaluating the Hidden Costs of Personalization in Large Language Models

  2608.28833 ▲ 28
• 52 Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching

  2609.01404 ▲ 27
• 53 Super Library Agent: Joint Generation and Maintenance of Multiple Applications Beyond the Single Codebase

  2608.29310 ▲ 27
• 54 The Attention Triangle in Audio-Video Models

  2609.03586 ▲ 25
• 55 DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training

  2609.04094 ▲ 25
• 56 CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation

  2609.04083 ▲ 25
• 57 ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes

  2609.01740 ▲ 25
• 58 NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference

  2609.01657 ▲ 25
• 59 WorldReward: Reward Modeling for Camera-Conditioned World Models

  2609.03952 ▲ 24
• 60 PACE: Towards Surfacing Hidden Conflicts in User Requests

  2609.03293 ▲ 23
• 61 Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System

  2609.01607 ▲ 23
• 62 WorldSculpt: Generating Compositional Worlds from Grounded Videos

  2609.05416 ▲ 21
• 63 DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory

  2609.00768 ▲ 20
• 64 Learning to Evaluate Before Improving: Automatic Rubric Induction for Automatic Research Agents

  2608.31076 ▲ 20
• 65 Safin-1: Safety from Within through Memory-Native State Evolution

  2609.00092 ▲ 20
• 66 FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow

  2609.03563 ▲ 19
• 67 Environment Evolution for Terminal Agents

  2609.04128 ▲ 19
• 68 Using Grounded Theory for Agent Behavior Analysis at Scale

  2608.30391 ▲ 19
• 69 Matrix-Game 3.5: Enhancing Real-Time Streaming Interactive World Models with Patch Memory

  2608.29910 ▲ 18
• 70 Cliff: Learning Process Rewards from the First Mistake

  2609.02817 ▲ 17
• 71 Enoki: Efficient Multi-Level Hallucination Detection

  2609.00581 ▲ 17
• 72 A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss

  2609.00591 ▲ 17
• 73 Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

  2609.01481 ▲ 17
• 74 Principia: Relational Physics Tests for Video Models

  2609.04200 ▲ 16
• 75 VibeVoice-ASR-Streaming Technical Report

  2609.02812 ▲ 16
• 76 E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation

  2608.30730 ▲ 16
• 77 Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation

  2608.29846 ▲ 16
• 78 Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

  2609.05258 ▲ 15
• 79 Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference

  2609.05275 ▲ 15
• 80 Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs

  2609.03820 ▲ 15
• 81 Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance

  2609.02373 ▲ 14
• 82 EM^2Mem: Event-Centric Multimodal Memory for Large Language Models

  2609.00551 ▲ 14
• 83 Lies We Can See: Joint Verbal and Non-Verbal Deception by VLM Agents in Embodied Social Interactions

  2608.30428 ▲ 14
• 84 QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation

  2608.29253 ▲ 14
• 85 UniMate: One Unified Model to Animate Diverse Skeletons

  2609.05415 ▲ 13
• 86 Let Confidence Change, Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration

  2609.01072 ▲ 13
• 87 Chain-of-Thought Faithfulness of Reasoning Models Varies with Where and How Preference Cues Are Delivered

  2608.29464 ▲ 13
• 88 Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation

  2608.24293 ▲ 13
• 89 VeriPhy: Agentic Physical Reasoning for World Model Evaluation and Refinement

  2609.03153 ▲ 12
• 90 PaperBanana-Interact: Scientific Diagram Refinement with Multi-Turn Human Feedback

  2608.30241 ▲ 12
• 91 Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers

  2608.18972 ▲ 11
• 92 Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs

  2609.04753 ▲ 10
• 93 MaxKernel: Agentic Kernel Generation for TPUs

  2609.04523 ▲ 10
• 94 A Common Measure of Communication for Speech Brain-Computer Interfaces

  2609.02887 ▲ 10
• 95 Agents in the Large: Perception-Centered Architecture for Persistent Agents

  2608.30478 ▲ 10
• 96 Verification-Aware Training for Speculative Decoding

  2608.30135 ▲ 10
• 97 Scaffolding Foundation Models into Physical-World Agents Pushes the Frontier of Long-Horizon Navigation

  2608.30396 ▲ 10
• 98 Locked at the Entrance, Open Inside: Where RLVR Narrows the Solution Space

  2608.29188 ▲ 10
• 99 RISE: Recursive Improvement via Self-Extrapolating Policy Distillation

  2609.05295 ▲ 9
• 100 Post-Training Language Models for Gold-Medal Performance in Coding Competitions

  2609.02849 ▲ 9
• 101 Learning Where Outcomes Change:Credit-Addressable Reasoning for Multimodal Geometry

  2608.30457 ▲ 9
• 102 CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing

  2609.01925 ▲ 8
• 103 Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall

  2609.01532 ▲ 8
• 104 MULTI3IR: A Benchmark for Multi-perspective Multi-domain Multi-modal Information Retrieval

  2608.30949 ▲ 8
• 105 WebWorld: The Browser as a World Model for Self-Improving Web Code

  2608.30530 ▲ 8