AI 论文 · 2026-07
浏览 2026-07 发布的AI 论文内容，第 1 页，共 585 条。
- Pushing the Limits of High-Resolution Weather Forecasting through Data Scaling
- DarwinX: Evolving Agent Harnesses Through Natural Selection
- Verifier-Induced Support Reshaping in On-Policy Optimization
- AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?
- DiffusionGemma Technical Report
- Toward Robust and 3D-Aware RGB-NIR Imaging in the Dark
- Scaling Properties of Text Conditioning in Visual Generation
- ExtractBench: A Benchmark for Schema-Guided Enterprise Document Extraction
- WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning
- Zero-Mem: Zero-Token Memory Operations for LLM Agents
- RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems
- Knowing When to Quit: Diagnosing and Training LLMs to Abort Futile Reasoning
- SAF-OPD: Stable Advantage Fusion for On-Policy Distillation
- A Frozen Pixel-Space Diffusion Model Can Guide Itself with Its Own Samples
- Evaluation-Verification Reward for Consistent Multi-Reference Image Editing
- SULAND v2: A Refined RGB Dataset and Deep Learning Object Detection Benchmark for UAV/UGV-Based SUrface LANDmine Detection Under Domain Shift
- ST-WAM: Semantic-Temporal World Action Model for Robust Manipulation under Visual Distribution Shifts
- MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations
- EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents
- To Add Is Machine, To Delete Is Human: Measuring and Mitigating Deletion Avoidance in LLM Code Editing
- Model or Harness? An Interaction-Centric Taxonomy for Localizing Agent Failures
- ReToken: One Token to Improve Vision-Language Models for Visual Retrieval
- ACE-Data-0: Human-Centric Ambient Capture as Embodied Data Engine
- PhiZero: A World Model Built Around Physical Language
- AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis
- AISPA: User-Centric System Prompt Auditing for Large Language Model Applications
- Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers
- OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
- Beacon: Knowing When and How to Perform Agentic Visual Reasoning
- VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation
- β-OPSD: Deriving with Policy Optimization, Training with Self-Distillation
- Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering
- RefCaptioner: Multi-Reference Image-Grounded Video Captioning
- Would You Walk to the Car Wash? Revealing the Salience Bias of Large Language Models in Commonsense Reasoning
- One Future, Every Robot: Label-Efficient Collective-State Prediction with Decentralized JEPA
- QQWorld: Quantile-Quantile Matching for World Model Regularization
- Can Large Language Models Execute Parent Orders?
- LEDGERMIND: Provenance-Constrained Multimodal Agentic Reasoning with a Structured Evidence Ledger
- ShadowDancer: Teaching Video World Models Any Action by Learning Unified Dynamics Representations from a Video and Its Shadow
- Fairness Pruning: Locating Demographic Bias in GLU-MLP Layers via Differential Activations
- Beyond Geometric Complementarity: Coherent Overlap in Sparse Mixture-of-Experts Routing
- MemHarness: Memory Is Reconstructed, Not Replayed
- Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
- Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale
- Flux-OPD: On-Policy Distillation with Evolving Contexts
- Σ-Mem: An Online Reliability Memory for LLM-based Multi-Agent Systems
- Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs
- Complementary Matrix-Gated QKAN Fast-Weight Programmers for Quantum Dynamics Forecasting
- ODEWorld: A Continuous Predictive Architecture via Physical-Time Flow
- Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory
- Not All Tokens Deserve Equal Credit: Counterfactual Sensitivity Credit Reallocation for Long-CoT Reasoning
- FinanceHarness: Autonomous Financial Deep Research Framework
- Beyond Feeling Better: Capability-Sustaining Emotional Dialogue as a Longitudinal Research Paradigm
- Beyond Borrowed Histories: Person-Aligned User Simulation for Interactive Role-Playing Evaluation
- Articulated Object Reconstruction from Rest-State Observation
- SpatialCLI: Learning to Reason With Spatial Tools, Then Without Them
- Harness-G: A Graph-Structured Harness for Search Agents
- MPIE-Bench: Benchmarking Anatomically Plausible Multi-Person Interaction Editing
- RL^2-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models
- BM25 Wins at Scale: A Scaling Study of Retrieval-Augmented Generation Paradigms
- LeapTalk: Breaking the Latency-Quality Trade-off in Talking Head Generation
- VideoCoCo: Code-as-CoT for Physically-Consistent Video Generation via an Agentic Dual-Engine System
- Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generation
- OVEarth-Bench: Evaluating Category Breadth and Query Diversity for Open-Vocabulary Earth Observation
- TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM
- Mental World Modeling
- Can AI agents conduct open-ended AI research? Early evidence from two case studies
- HumanCLAW: Can Vision-Language Models Act Through a Body?
- SpecFirst: Behavioral Specification Elicitation as a First-Class Step in Agent-Based Program Synthesis from Scratch
- OmegaUse-OfficeVal: Benchmarking LLM Agents on Long-Horizon Office-Suite Tasks with Economic Grounding
- MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis
- GPTQ-2D: Cubic-Time Two-Sided Adaptive Rounding
- ICDAR 2026 Competition on Information Extraction from Atomic Layer Deposition/Etching (ALD/E) Scientific Figures
- DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillation
- SecRespond: Benchmarking AI Agents for Real-World Post-Compromise Incident Response
- SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution
- See2Think: Do Multimodal Models Really Use Intermediate Visual States?
- Metis: Memory Foundation Model
- StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation
- Constitutional Midtraining: Content Presence Drives Alignment Gains
- Filesystem-Based Memory for LLM Agents: Organization, Evolution, and Sustainability
- Revisiting Lossy Verification in Speculative Decoding: Mechanisms, Trade-offs, and Failure Modes
- Fewer Clarifications, Better Code: Benchmarking Cross-Session Personalized Ambiguity Adaptation in Coding Assistants
- ExplainBench: Evaluating Code Explanations from Agents
- Voice Memory for Agentic Speech Recognition
- Valid Per-Field Selective Risk Control for Document Extraction: Three Failure Modes, a Validity Ladder, and When Conditioning Pays
- Meshy T2: Fast Native Mesh Generation with Flow Matching
- Seeing or Knowing? Visual Context Sensitivity in Multimodal Large Language Models
- StealthBench: Measuring Operational Stealth in Autonomous Offensive-Security Agents
- Weak-to-Strong On-Policy Distillation
- DuplexGen: Adaptive Synthesis of Human-AI Turn-Taking Dialogues
- GPT-Red: Automated Red Teaming via Self-Play at Scale
- Pass the Baton: Trajectory-Relayed On-Policy Distillation
- INTACT: Isomorphic Intent-to-Action Learning for Search-Free World Models
- πR^2: Reactive Real-time Flow Policies
- Wonder: Video World Model Done Better
- Parallel Decoding Distillation for Fast Image and Video Generation
- Reinforcement Learning for Code Optimization
- MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities
- HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone