AI 论文 · 2026-09
浏览 2026-09 发布的AI 论文内容，第 3 页，共 741 条。
- AgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents
- Allspark: Weak to Strong Transfer via Alternating Chain of Thought
- Change the Product, Keep the Parameters: Associative Algebra Layers for Transformers
- OpenTumorBoard: A Real-World Benchmark of Multidisciplinary Tumor Board Discussion Trajectories
- The Extender: A Log-Structured Transformer
- Scaling Properties of Same-Family On-Policy Distillation
- BiasReducer: Adaptive Bias Mitigation for Reward Models
- ExpVoyager: Direct Experience Navigation for Dynamic Agent Skill Synthesis
- VoxMem: Benchmarking Multimodal Memory in Large Audio Language Models
- CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering
- Groupwise Agentic Grading and Advantage Redistribution for Code Agent RL
- In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion
- Beyond Dyadic Memory: Interaction-Aware Multimodal Memory with Adaptive Agentic Retrieval for Multi-Party Spoken Conversations
- GeoCR: Learning a Generalist Cloud Removal Prior from Heterogeneous Observations
- LANTERN: Illuminating Hidden Mathematical Knowledge in Language Models
- Residual Transferability in Neural Image Watermarking
- KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding
- Playing to Par: Reinforcement Learning for Provably Optimal Quadrilateral Block Decompositions
- Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration
- Equal Ranking Quality, Different Decisions: Measuring and Reducing Order Dependence in LLM Scorers
- Mara Chain: Rethinking Failure as a Stepping Stone for AI System Auto-Evolution
- Hyperspherical Semantic Trajectory Analysis: Mapping Technological Diffusion across Academic Preprints, Patent Signals, and Compute Scaling
- Receiver-Conditioned Latent Communication gives 94% CacheBack
- Decentralized Master-Mind: Joint Action Refinement through Iterative Intent Denoising in Multi-Agent Pathfinding
- NVAlign: Direct-Gradient Optimization for Non-Verbal Control in Continuous Autoregressive Flow Matching Text-to-Speech
- Omni-IO Skills: Harnessing Your Agent Omni-Native
- FuseReg: Regularizing Layer Fusion Mitigates the Reconstruction-Generation Gap in Representation Autoencoders
- AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs
- SatNav: A Scalable Benchmark for Long-Horizon UAV Vision-Language Navigation from Satellite Imagery
- Game Arena: Strategic LLM Evaluation in Competitive Environments
- InternW0-Δ: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data
- Softmax Reparameterization for Output-Head Quantization
- Enhancing Photogrammetric Digital Surface Models with Pretrained Diffusion Models and Multimodal Conditioning
- Block Sparse Attention with Log-Linear Complexity
- G^2PTQ: Improving LLM Post-Training Quantization with Generalized Gradient Compensation
- ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker
- Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations
- Aligning One-Step Generative Models with Reward-Weighted Transport Distillation
- MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation
- TRACE: Temporal Audit and Condition-aware Evaluation of Streaming Video Understanding
- JumpStart Your Policy Learning with Lessons from 160,000 Training Runs
- Depth-adaptive Inference of Looped Language Models via Continuous Depth Batching
- Synthesis Through Simulation: Generating Coherent Enterprise Data via Scalable Agent-System Interaction
- AutoResearch at Production Scale: Failure Modes and a Multi-Agent Framework
- Where Does Retrieval-Based Open-Ended Evaluation Fail? Automatic Taxonomy Induction from Long-Form Medical Answer Factuality Verification
- Coding Agents for Generalized Task and Motion Planning Problems
- TrackEverything: Long Horizon Dense Tracking via De-Duplicating 3D Scene Representations
- WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation
- Jev in the Wild: A Data-Driven Analysis of the Jev Model's Functionality, Applications and Ecosystem
- ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds
- SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance
- Rate-distortion optimization for full-reference image quality metrics via stochastic Hessian estimates
- How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure
- World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal
- Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents
- Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs
- PUBG Ally: A Conversational Embodied Agent as an AI Teammate
- AV-GRPO: Modality-Anchored Decoupling Diffusion Reinforcement Learning for Joint Audio-Video Generation
- TimeBraid: Unifying Time Series and Language for Understanding and Forecasting
- CodeGraph: Open-Taxonomy Knowledge Graph for Source Code with Wikidata Grounding
- IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis
- Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures
- Rufus-Air: An Open LLM Post-Training Recipe
- Parts-of-Speech as Emergent Categories in SAE Latent Space
- IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking
- Accent Analogy Guidance: More Speaker Similarity at Equal Accent in Cross-Lingual Voice Cloning
- SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL
- RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation
- ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generation
- LightMIS: Ultra-Lightweight Medical Image Segmentation Without a Stage-Wise Decoder
- Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents
- LastOPD: Taming Collapse in Latent On-Policy Distillation
- DeltaWAM: Delta World Action Models for Bimanual Manipulation
- Morphometric Imitation: From Morphology and Contact Aware Hand Retargeting to Sim-to-Real Visuomotor Policy
- Training Object Permanence in World Models
- Learning to Discover Interesting Mathematics
- On the Diffusibility of High-Dimensional Latents
- StudentBench: AI and human tutoring yield equivalent GRE learning gains
- The Past Frames the Future: Memory for Autoregressive Video Generation
- Agent-Editing World Model: Rethinking World Modeling for LLM Agents
- Beyond Future Prediction: Denoising as Generative Adaptation for Robot Control
- MemBodied: Recurrent Associative Memory for Vision-Language-Action Models
- Six Layers Less: Encoder Pruning for Whisper with Label-Free Recovery
- All modalities are equal, but video is more equal: Closing the Cross-Attention Gap in Joint Video Generation
- FLEET: From Logits Entropy to Enhanced Trajectories in Text Generation
- InternW0: A Foundational Physical World Model for Efficient Real-World Interactions
- WhatWorkedBench: Benchmarking Experimental Understanding in AI Agents
- Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents
- Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms
- EmbodiedSWE: Coding Agents for Long Horizon Dexterous Robotics
- Hunyuan-A13B Technical Report
- TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent
- Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI
- The Linear Representation Hypothesis Needs a Group Action
- Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
- HARMONY: Hierarchical Agentic Reasoning for MONocular Image-to-Scene Synthesis
- Agensh: Scaling Organizational Intelligence to 1,024 Agents
- SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue
- StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training
- Capable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models