AI 论文 · 2026-09
浏览 2026-09 发布的AI 论文内容，第 4 页，共 741 条。
- Knowledge Pull Requests for Continual Document Authoring
- JEV-as-a-Judge: Accept When Confident, Escalate When Unsure
- Calibration as a First-Class Criterion in LLM Evaluation
- Recursive self-improvement of AI research agents
- PACT: From Credit Assignment to Critic Alignment
- Blaming Across the Aisle: Political Contrasting and Blame Attribution in the Danish Parliament
- Disaggregated Quantization: Specializing LLM Prefill and Decode
- GeoPair: Geometry-Preserving Cross-Layer Factorization for Training-Free Transformer Compression
- MemoryAthena: Adaptive Routing over Latent and Generated Memories
- The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks
- FoMo: Forking Moment in Generative Trajectory as a Perceptual Distance
- RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents
- D-JEPA: A Decision-Aligned Latent World Model
- RULER: Instance-aware Rubric Rewards for SVG Generation
- ImIR: Image-Instruction Tuning for All-in-One Image Restoration
- Geometric and Semantic Coupling for Interaction Understanding in 3D Scenes
- Lean Pool: An AI-Maintained Archive of Formalized Mathematics
- X-Planner: Event-Structured Task Planning for Embodied Intelligence
- From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health
- Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings
- GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay
- VideoGen-Agent: Reinforcing Video Generation Agents
- WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory
- onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction
- GAE: Learning a Geometry-Native Latent Space for 3D-Consistent World Generation
- Harness-Zero: Harness Distillation via Agent-as-Harness
- RRSI: Regularized Recursive Self-Improvement of Agent Harnesses
- Emergent Collusion in Long-Horizon LLM Agent Interaction
- Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention
- Streaming Video Editing with Easy Adaptation
- Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies
- Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models
- 1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation
- Tactile-JEPA: Topology-Aware Self-Supervised Representation Learning for Distributed Tactile Sensors
- HappyWorld-Bench
- Document Retrieval-Aware Chunking (D-RAC): Universal Retrieval-Aware Ingestion of Enterprise Documents via PDF Normalization and Multimodal Markdown Conversion
- CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies
- EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation
- All-in-One Multilingual Scene Text Recognition with Script-aware Mixture-of-Experts
- ACLArena: Agent Continue Learning in Multi-stage Post-training
- Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents
- Not All Objectives Are Born Equal: Priority-Constrained Descent for Hierarchical Multi-Objective Optimization
- Grounded Action Model: 3D Grounding as a Foundation for Robotics
- Mira-Scene: Pixel-Aligned Layouts for Generative 3D Scene
- PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing
- Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms
- Paragraph Boundaries Are Not White Space:Compression Depth as the Signature of Hierarchical Structure
- OmniEcho: Spatial Audio Understanding for Embodied Agents
- One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents
- Specification Before Generation: A Pre-Registered, Five-Model Paired Evaluation of a Specification Frame for LLM-Generated Code in Money, Time, Idempotency, and Access Tasks
- BoundInk: Boundary-Aware Online Handwriting Generation
- Do Implicit Personalization and Explicit Styles Conflict? PsPLUG: A Lightweight Plug-in for Balancing Personalization and Style in Customized LLMs
- CARD: Cluster-level Adaptation with Reward-guided Decoding for Personalized Text Generation
- UltraTex: Unleashing 2K Multi-View Diffusion for 3D Texturing
- OmniEdu: Open Foundation Models for Learning and Teaching
- Neural Spectral Capacity: Measuring and Designing Architectures from Network Specification Alone
- Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World
- Transferring the Intelligence of VLMs to Robotic Control
- RewardVerse: Rubric-Guided Policy Optimization for Video Reward Modeling
- Towards Full Pipeline FP8 Reinforcement Learning for LLMs
- Self-Organizing Agent Teams Learn to Reason Together
- Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design
- MintAct: A Unified Visual Agent for Digital Environments
- APort Vault: Benchmarking AI Agent Payment Authorization with the Open Agent Passport
- OmniVBench: A Benchmark and Large-Scale Dataset for Omni Reference-to-Video Generation
- CodeMidas: Scaling Agentic Coding RL Environments from Code Itself
- Gricea: An Open Science Platform for Conversational AI Research
- RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents
- A Lie Detector Test for Language Models: Reading Knowledge a Model Won't Reveal
- From Pretraining to Proficiency: Real-World Subtask RL for Long-Horizon Manipulation with Minimal Human Intervention
- GraphSkillEvo: Evolutionary Optimization of Graph-Structured Agent Skills
- Calibrating Teacher--Student Discrepancy for On-Policy Distillation
- OmniVChat: Synthesizing, Benchmarking, and Training for Native Audio-Visual Dialogue
- IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation Mixture-of-Experts
- The Functionalizer: Lossless Functional Decomposition for Subword Tokenization
- HuRo: Robotizing Human Videos for Scalable VLA Pretraining
- VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models
- Geometry of Values: Task Vector Composition for Ethical Preference Alignment in Language Models
- Retention-Constrained Post-Training Quantization of Cellpose-SAM for Stem Cell Microscopy
- When AI Reviews Train AI Reviewers: Scientific-Judgment Collapse and Mitigation
- FAMOS: Feed-Forward 3D Articulation Modeling from Sparse Observations
- Paint-Anything: Unified Any-Color Control for Image Generation and Editing
- An Empirical Study of Harness Design for Coding Agents
- JEPA-Anything: Learning Predictive Models across Different Worlds
- RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning
- Prediction-Powered Smoothing and Validation for Disaggregated AI Evaluation
- Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video Generation
- Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
- Learning Foresight without Explicit Trajectories for 3D Diffusion Policies
- Refinement Is Inherently Editable: Training-Free Prompt-to-Prompt Image Editing with Generative Refinement Network
- What Does Privileged Information Add to On-Policy Self-Distillation?
- SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness
- When EOS Tokens Disagree: Understanding Length Inflation in On-Policy Distillation
- WeVisDoc: From Coverage to Capability for Robust End-to-End Document Parsing
- DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
- VākQA: A Benchmark and Evaluation Study for Telugu Spoken Factoid Question Answering
- Region-Level Policy Optimization for Fine-grained MLLM Perception
- When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models
- Self-Evolving Search Index
- VABench: Measuring Embodied Spatial Intelligence through Visual Demonstrations, Active Perception, and Metric Control