每日论文 Daily Papers

来自 Hugging Face 与研究社区 · 中英对照

Daily Weekly Monthly

‹ 2026-06 ›

跳转日期： 09-11 30 09-09 10 09-08 26 09-07 31 09-04 31 09-03 26 08-30 2 08-29 21 08-28 32 08-27 23 08-26 32 08-25 21 08-23 11 08-21 33 08-19 59 08-18 31 08-15 30 08-14 23 08-11 30 08-07 47 08-06 36 08-05 37 08-04 25 08-02 38 07-31 17 07-30 23 07-29 27 07-28 16 07-25 21 07-24 19

JoyAI-VL-Interaction: Real-Time Vision-Language Interaction Intelligence

▲ 157 💬 2 ★ 209

Moebius：0.2十亿像素的轻量级图像修复框架，性能达到100亿像素级别 Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance

▲ 102 💬 4 ★ 66

Data Journalist Agent: Transforming Data into Verifiable Multimodal Stories

▲ 100 💬 4 ★ 45

Geometric Action Model for Robot Policy Learning

▲ 85 💬 3 ★ 56

我们准备好迎接一种基于代理的记忆系统了吗？ Are We Ready For An Agent-Native Memory System?

▲ 83 💬 3 ★ 20

Qwen-AgentWorld：通用智能体语言世界模型 Qwen-AgentWorld: Language World Models for General Agents

▲ 79 💬 1

DreamX-World 1.0：一种通用的交互式世界模型 DreamX-World 1.0: A General-Purpose Interactive World Model

▲ 78 💬 7 ★ 318

PlanBench-XL：评估大型工具生态系统中LLM工具使用智能体在长期规划方面的表现 PlanBench-XL: Evaluating Long-Horizon Planning of LLM Tool-Use Agents in Large-Scale Tool Ecosystems

▲ 78 💬 3 ★ 26

FastContext：训练出高效的海量代码代理存储查询工具 FastContext: Training Efficient Repository Explorer for Coding Agents

▲ 76 💬 2 ★ 407

OpenRath：以会话为中心的代理系统运行时状态 OpenRath: Session-Centered Runtime State for Agent Systems

▲ 68 💬 3 ★ 672

DataClaw0：从原始数据流中对多模态数据进行智能处理 DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams

▲ 65 💬 1 ★ 4

DragMesh-2：具有物理真实性的灵巧手与可关节式物体的交互 DragMesh-2: Physically Plausible Dexterous Hand-Object Interaction with Articulated Objects

▲ 64 💬 4 ★ 5

VibeThinker-3B：探索小型语言模型中可验证推理的边界 VibeThinker-3B: Exploring the Frontier of Verifiable Reasoning in Small Language Models

▲ 62 💬 1 ★ 597

EnterpriseClawBench：基于真实工作场景的代理对比分析 EnterpriseClawBench: Benchmarking Agents from Real Workplace Sessions

▲ 56 💬 1 ★ 28

DomainShuttle：自由形式的开放领域主题驱动文本到视频生成 DomainShuttle: Freeform Open Domain Subject-driven Text-to-video Generation

▲ 55 💬 1 ★ 66

DanceOPD：策略内生成式字段蒸馏 DanceOPD: On-Policy Generative Field Distillation

▲ 52 💬 2

PerceptionDLM：基于多模态扩散语言模型的并行区域感知技术 PerceptionDLM: Parallel Region Perception with Multimodal Diffusion Language Models

▲ 49 💬 4 ★ 24

NatureBench：编码智能体能否达到Nature系列论文中已发表的最先进水平？ NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers?

▲ 47 💬 1 ★ 29

分组查询专家：GQA自注意力中的专家混合模型 Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention

▲ 42 💬 1

ACE-Ego-0：将以自我为中心的人类数据与机器人数据整合，用于VLA预训练 ACE-Ego-0: Unifying Egocentric Human and Robotic Data for VLA Pretraining

▲ 40 💬 2 ★ 10

KaLM-Reranker-V1：用于压缩文档重新排序的快速但非延迟交互方法 KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking

▲ 39 💬 2

近似策略优化：以教师为主导的优化过程，而非通过梯度来优化策略 Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients

▲ 38 💬 2

Wan-Streamer v0.1：端到端实时交互基础模型 Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models

▲ 38 💬 2

PhysisForcing：用于机器人操作的物理强化世界模拟器 PhysisForcing: Physics Reinforced World Simulator for Robotic Manipulation

▲ 38

有趣的主体式机器人学习 Playful Agentic Robot Learning

▲ 37 💬 2 ★ 38

世界行动模式：一项调查 World Action Models: A Survey

▲ 37 💬 1 ★ 233

超越当前观察：在可控非马尔可夫游戏中评估多模态大语言模型 Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games

▲ 36 💬 5 ★ 33

ShutterMuse：基于MLLMs的拍摄时间摄影指导工具 ShutterMuse: Capture-Time Photography Guidance with MLLMs

▲ 36 💬 1 ★ 11

GameCraft-Bench：代理程序能否在真实的游戏引擎中从开始到结束完整地构建出可玩的游戏？ GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?

▲ 35 💬 2 ★ 44

Multi-LCB：将LiveCodeBench扩展至多种编程语言 Multi-LCB: Extending LiveCodeBench to Multiple Programming Languages

▲ 34 💬 3 ★ 22

MobileForge：无需注释的移动GUI代理自适应方法，采用分层反馈引导的策略优化机制 MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization

▲ 34 💬 1 ★ 49

ViQ：任意分辨率的文本对齐视觉量化表示 ViQ: Text-Aligned Visual Quantized Representations at Any Resolution

▲ 34 💬 1 ★ 36

MemGUI-Agent：一种具备主动上下文管理的端到端长距离移动式GUI代理工具 MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management

▲ 33 💬 1 ★ 53

OPID：基于策略的技能蒸馏在代理式强化学习中的应用 OPID: On-Policy Skill Distillation for Agentic Reinforcement Learning

▲ 32 💬 1 ★ 26

翻译作为一种连接手段：将人类的操作技能转移给机器人 Translation as a Bridging Action: Transferring Manipulation Skills from Humans to Robots

▲ 32 💬 1

S-Agent：空间工具使用能够激发与空间智能相关的推理能力 S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence

▲ 31 💬 2 ★ 26

Qwen-Image-Agent：填补现实世界图像生成中的上下文空白 Qwen-Image-Agent: Bridging the Context Gap in Real-World Image Generation

▲ 31

LectūraAgents：一种多智能体框架，旨在实现自适应化的个性化AI辅助学习和基于实体的教学体验。 LectūraAgents: A Multi-Agent Framework for Adaptive Personalized AI-Assisted Learning and Embodied Teaching

▲ 30 💬 1

验证视野：编码助手奖励中没有“灵丹妙药” The Verification Horizon: No Silver Bullet for Coding Agent Rewards

▲ 30 💬 3

CLI-Universe：面向终端代理的可验证任务合成引擎 CLI-Universe: Towards Verifiable Task Synthesis Engine for Terminal Agents

▲ 27 💬 1

超越NL2Code：多模态代码智能的结构化调查 Beyond NL2Code: A Structured Survey of Multimodal Code Intelligence

▲ 27 💬 1 ★ 263

改进的大型语言扩散模型 Improved Large Language Diffusion Models

▲ 26 💬 1

分类：用于解释性风险预测的辩证推理方法——适用于具有不规律采样数据的医学时间序列数据 TRIAGE: Dialectical Reasoning for Explainable Risk Prediction on Irregularly Sampled Medical Time Series with LLMs

▲ 25 💬 2 ★ 1

Kairos：一种用于物理人工智能的本地世界模型栈 Kairos: A Native World Model Stack for Physical AI

▲ 25 💬 2 ★ 730

DF3DV-1K：一种用于无干扰新视角生成的大规模数据集与基准测试集 DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis

▲ 25 💬 3 ★ 20

AOHP：一种开源的操作系统级代理平台，用于实现个性化、高效且安全的交互体验 AOHP: An Open-Source OS-Level Agent Harness for Personalized, Efficient and Secure Interaction

▲ 25 💬 2 ★ 63

Qwen-Image-2.0-RL技术报告 Qwen-Image-2.0-RL Technical Report

▲ 25 💬 3

BRDFusion：物理原理与新一代城市场景反渲染技术的结合 BRDFusion: Physics Meets Generation for Urban Scene Inverse Rendering

▲ 24 💬 1 ★ 21

MVTrack4Gen：作为几何监督的多视图点跟踪技术，用于4D视频生成 MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video Generation

▲ 24 💬 1 ★ 20

现在，谁应该负责解码任务呢？我们需要找到可靠的路径来整合那些被隐藏起来的扩散语言模型。 Who Should Lead Decoding Now? Tracking Reliable Trajectories for Ensembling Masked Diffusion Language Models

▲ 23 💬 1

从“自我未来”中学习：适用于多语言大语言模型的自策略自融合方法 Learning from the Self-future: On-policy Self-distillation for dLLMs

▲ 23 💬 1 ★ 1

超越静态排行榜：用于评估LLM智能体的预测有效性 Beyond Static Leaderboards: Predictive Validity for the Evaluation of LLM Agents

▲ 23 💬 1

EvoEmbedding：用于长上下文检索与智能体式记忆的可进化表示方法 EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory

▲ 23 💬 3 ★ 21

将潜在思维形式化：大语言模型中的四种思维表示公理 Formalizing Latent Thoughts: Four Axioms of Thought Representation in LLMs

▲ 23 💬 4 ★ 1

番石榴：一种有效且通用的用于实体操作的工具 Guava: An Effective and Universal Harness for Embodied Manipulation

▲ 22 💬 4

FreeStyle：通过社区LoRA挖掘技术实现对风格与内容双重参考的生成方式的自由控制 FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining

▲ 22 💬 1 ★ 10

VisualClaw：一种适用于物理世界的实时、个性化智能代理 VisualClaw: A Real-Time, Personalized Agent for the Physical World

▲ 21 ★ 15

OPD-Evolver：通过基于策略的蒸馏方法，培育出具有全面能力的智能体 OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillation

▲ 20 💬 1 ★ 7

BioMatrix：构建涵盖序列、结构和语言各维度的综合生物基础模型 BioMatrix: Towards a Comprehensive Biological Foundation Model Spanning the Modality Matrix of Sequences, Structures, and Language

▲ 19 💬 1 ★ 19

LingxiDiagBench：一种用于比较中文精神科咨询与诊断领域大型语言模型性能的多智能体框架 LingxiDiagBench: A Multi-Agent Framework for Benchmarking LLMs in Chinese Psychiatric Consultation and Diagnosis

▲ 19 💬 1 ★ 9

V-Zero：基于对比性证据门控的无需答案标签的在线策略蒸馏方法，用于精细视觉推理 V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning

▲ 19 💬 1 ★ 14

UnityShots：基于内存的多帧音视频生成技术，具有边界感知门控功能 UnityShots: Memory-Driven Multi-Shot Audio-Video Generation with Boundary-Aware Gating

▲ 19 💬 1 ★ 2

JetSpec：通过并行树绘制技术突破投机性解码的扩展限制 JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting

▲ 19 💬 1 ★ 35

JanusMesh：通过跨空间去噪技术实现快速且无需样本训练的3D视觉幻象生成 JanusMesh: Fast and Zero-Shot 3D Visual Illusion Generation via Cross-Space Denoising

▲ 17 💬 2 ★ 14

HydraHead：从头部级别的功能异质性到专门化注意力的混合化 HydraHead: From Head-Level Functional Heterogeneity to Specialized Attention Hybridization

▲ 17 💬 1

FlowBender：针对自修正条件流的基于反馈的训练方法 FlowBender: Feedback-Aware Training for Self-Correcting Conditional Flows

▲ 16 💬 2

OpenThoughts-Agent：代理式模型的数据配方 OpenThoughts-Agent: Data Recipes for Agentic Models

▲ 16

MultiHashFormer：基于哈希的生成式语言模型 MultiHashFormer: Hash-based Generative Language Models

▲ 16 💬 1 ★ 1

显示信号，隐藏噪声：用于像素空间扩散的谱激励方法 Show the Signal, Hide the Noise: Spectral Forcing for Pixel-Space Diffusion

▲ 15 💬 1 ★ 11

SkillHarness：利用安全技能来辅助计算机使用智能体 SkillHarness: Harnessing Safe Skills for Computer-Use Agents

▲ 15 💬 1 ★ 1

OneRank：一种统一的Transformer与原生排序架构，适用于多任务推荐场景 OneRank: Unified Transformer-Native Ranking Architecture for Multi-Task Recommendation

▲ 14 💬 1

BadWorld：对世界模型的对抗性攻击 BadWorld: Adversarial Attacks on World Models

▲ 14 💬 1 ★ 7

Qwen-RobotWorld技术报告：通过语言驱动的视频生成来实现具身世界的统一建模 Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation

▲ 14 💬 1

EfficientRollout：基于系统感知的自猜测译技术，用于强化学习部署 EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts

▲ 14 💬 3 ★ 1

SAE干预不可靠：干预后被抑制行为的恢复情况 SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

▲ 14 💬 3 ★ 8

从自己的错误中学习：构建可学习的微型反思路径以实现自我反思 Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation

▲ 14 💬 1

SP^3：适用于即插即用的球形修复方法 SP^3: Spherical Priors for Plug-and-Play Restoration

▲ 13 💬 1 ★ 2

TokenPilot：适用于大语言模型智能体的高效缓存式上下文管理方案 TokenPilot: Cache-Efficient Context Management for LLM Agents

▲ 13 💬 1 ★ 15

在空间视觉语言模型中强化双路径推理 Reinforcing Dual-Path Reasoning in Spatial Vision Language Models

▲ 13 💬 1 ★ 4

相信合适的老师：注重质量的自我蒸馏法用于GUI grounded Trust the Right Teacher: Quality-Aware Self-Distillation for GUI Grounding

▲ 13 💬 2

GateMem：多主共享内存代理中的内存管理基准测试 GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents

▲ 13 💬 1 ★ 74

MemSlides：一种基于分层内存的智能体框架，用于生成个性化幻灯片，并支持多轮本地修改功能 MemSlides: A Hierarchical Memory Driven Agent Framework for Personalized Slide Generation with Multi-turn Local Revision

▲ 13 💬 2 ★ 4

IV-CoT：用于结构感知文本到图像生成的可隐式视觉思维链 IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation

▲ 13 💬 1

EBench：通用移动操作策略的要素诊断 EBench: Elemental Diagnosis of Generalist Mobile Manipulation Policies

▲ 13 💬 1 ★ 95

无限OCR功能 Unlimited OCR Works

▲ 12 ★ 2570

更深并不总是更好：通过自信的层解码来降低对齐成本 Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding

▲ 12 💬 7 ★ 5

物理科学深度研究：多智能体框架与全面基准测试 Deep Research in Physical Sciences: A Multi-Agent Framework and Comprehensive Benchmark

▲ 12 💬 1 ★ 2

FLAT：用于几何精度场景生成的前馈隐式三角形斑点算法 FLAT: Feedforward Latent Triangle Splatting for Geometrically Accurate Scene Generation

▲ 12 💬 1

语义浏览：可控制的图像生成多样性 Semantic Browsing: Controllable Diversity for Image Generation

▲ 12 💬 1

CODA-BENCH：代码代理能够处理那些需要大量数据处理的任务吗？ CODA-BENCH: Can Code Agents Handle Data-Intensive Tasks?

▲ 11 💬 1 ★ 6

提醒：通过重构来记住那些能够持续生成长视频的内容。 Memento: Reconstruct to Remember for Consistent Long Video Generation

▲ 11 💬 2 ★ 6

重新思考高效注意力在混合架构中的角色 Rethinking the Role of Efficient Attention in Hybrid Architectures

▲ 11 💬 1 ★ 1

从培训生到培训师：基于LLM设计的、支持多智能体推理的强化学习培训环境 From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning

▲ 11 💬 1 ★ 7

预见性：基于动作条件世界模型隐式的长距离机器人操作故障检测 Foresight: Failure Detection for Long-Horizon Robotic Manipulation with Action-Conditioned World Model Latents

▲ 11 💬 1

因果-RCM：一种统一的教师强制与自我强制方法，用于流式视频生成和交互式世界模型中自回归扩散蒸馏 Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models

▲ 11

SingGuard：一种具有动态推理能力的、能适应政策的多模态LLM监管框架 SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning

▲ 11 💬 1 ★ 30

无需重新训练：在测试阶段即可将视觉语言动作模型应用于新的任务中。 Retrieve, Don't Retrain: Extending Vision Language Action Models to New Tasks at Test Time

▲ 10 💬 1

问题出在哪儿？基于语义状态的Web代理过程级评估 Where Did It Go Wrong? Process-Level Evaluation of Web Agents with Semantic State Tracking

▲ 10 💬 1

文本与视觉结合的图像编辑方式 Text-Vision Co-Instructed Image Editing

▲ 10 💬 3 ★ 14

作为全模态理解基础的本土主动感知机制 Native Active Perception as Reasoning for Omni-Modal Understanding

▲ 10 💬 1 ★ 14

ImageWAM：世界行动模型真的需要视频生成吗？还是只是需要图像编辑？ ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?

▲ 10 ★ 27

自压实语言模型智能体 Self-Compacting Language Model Agents

▲ 10

为什么多步骤工具使用强化学习机制会崩溃，以及监督信号如何解决这一问题 Why Multi-Step Tool-Use Reinforcement Learning Collapses and How Supervisory Signals Fix It

▲ 10 💬 1 ★ 2

Fast LeWorldModel

▲ 10 ★ 2

GD^2PO：通过群体动态奖励分离式策略优化来缓解多奖励冲突问题 GD^2PO: Mitigating Multi-Reward Conflicts via Group-Dynamic reward-Decoupled Policy Optimization

▲ 9 💬 1 ★ 7

Ling和Ring 2.6技术报告：在万亿参数规模下实现高效且实时的智能代理系统 Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale

▲ 9 💬 2

自我演化视觉提问器 Self-Evolving Visual Questioner

▲ 9 💬 1

当前的世界模型缺乏一个稳定的核心状态 Current World Models Lack a Persistent State Core

▲ 9 💬 1

Notes2Skills：从实验笔记到具备确定性意识的科学智能体技能 Notes2Skills: From Lab Notebooks to Certainty-Aware Scientific Agent Skills

▲ 9 💬 1

FedOT：通过水印技术对所有权进行验证并追踪数据泄露情况，适用于联邦式LDM系统 FedOT: Ownership Verification and Leakage Tracing via Watermarks for Federated LDMs

▲ 9 💬 1

《搭车客指南：代理式人工智能指南：从基础到系统》 The Hitchhiker's Guide to Agentic AI: From Foundations to Systems

▲ 9

挑战之旅：重新评估非熟悉环境中的代理人能力 Running the Gauntlet: Re-evaluating the Capabilities of Agents Beyond Familiar Environments

▲ 9 💬 1 ★ 1

唐格拉姆：解锁非均匀KV缓存压缩技术，以实现高效的多轮次LLM服务 Tangram: Unlocking Non-Uniform KV Cache Compression for Efficient Multi-turn LLM Serving

▲ 8 💬 2 ★ 6

UniDDT：通过分离扩散变换技术，实现多模态理解与生成的统一处理 UniDDT: Unifying Multimodal Understanding and Generation with Decoupled Diffusion Transformer

▲ 8 💬 1

PhoneHarness：通过混合式GUI、命令行界面以及工具操作来优化手机使用体验 PhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions

▲ 8 💬 1 ★ 23

Nemotron 3 Ultra：一种开放的、高效的混合专家模型，适用于代理推理任务 Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning

▲ 8 💬 1

ActWorld: From Explorable to Interactive World Model via Action-Aware Memory

▲ 8 圣路易斯华盛顿大学、字节跳动 · 计算机视觉、世界模型、视频生成、交互式环境模拟

MolmoMotion：通过语言教学进行3D中点轨迹预测 MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction

▲ 8 💬 1 ★ 35

STARE：基于惊喜机制的令牌级优势重新加权方法，用于确保策略熵值的稳定性 STARE: Surprisal-Guided Token-Level Advantage Reweighting for Policy Entropy Stability

▲ 8 💬 1 ★ 13

每日报告：用于评估搜索代理在日常搜索任务中的表现的一种开放式基准标准 DailyReport: An Open-ended Benchmark for Evaluating Search Agents on Daily Search Tasks

▲ 8 💬 1 ★ 5

摆脱自我确认陷阱：一种用于代理式体验学习的执行-蒸馏-验证模式 Escaping the Self-Confirmation Trap: An Execute-Distill-Verify Paradigm for Agentic Experience Learning

▲ 8 💬 2

LISA：用于可控制生成效果的视觉条件概率评分对齐 LISA: Likelihood Score Alignment for Visual-condition Controllable Generation

▲ 8 💬 1 ★ 4

低资源语言文本净化系统Tatoxa：以鞑靼语为例 The Tatoxa System for Text Detoxification in Low-Resource Languages: The Case of Tatar

▲ 8 💬 1

EgoCS-400K：用于世界模型的自中心性游戏数据集 EgoCS-400K: An Egocentric Gameplay Dataset for World Models

▲ 7 💬 1 ★ 16

Sumi：从零开始构建统一扩散语言模型 Sumi: Open Uniform Diffusion Language Model from Scratch

▲ 7 ★ 6

MaineCoon：构建实时音视频社交世界模型 MaineCoon: Pursuing A Real-Time Audio-Visual Social World Model

▲ 7 💬 1 ★ 24

超越一致性：价值多样性作为多元文化主体系统中的集体财产 Beyond Alignment: Value Diversity as a Collective Property in Multicultural Agent Systems

▲ 7 💬 1 ★ 2

ENPIRE：现实世界中基于代理的机器人政策自我改进 ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

▲ 7

BrainG3N：一种用于可控3D脑MRI生成的双用途分词器 BrainG3N: A Dual-Purpose Tokenizer for Controllable 3D Brain MRI Generation

▲ 7 💬 1

Tmax：终端药物的一种简单制备方法 Tmax: A simple recipe for terminal agents

▲ 7 ★ 73

训练用于智能电话使用的开放模型 Training Open Models for Agentic Phone Use

▲ 7 💬 1 ★ 5

通过速度编辑实现安全的少步生成 Safe Few-Step Generation via Velocity Editing

▲ 7 💬 1

PoLAR：机器人策略学习中潜在动作的因子分解与模式识别 PoLAR: Factorizing Extent and Mode in Latent Actions for Robot Policy Learning

▲ 7 💬 1 ★ 1

连接点：通过强化学习训练长生命周期智能体，实现跨领域泛化能力 Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning

▲ 7 💬 1

探索奖励反向传播在流匹配中的设计空间 Exploring the Design Space of Reward Backpropagation for Flow Matching

▲ 7 💬 1 ★ 8

Autodata：一种智能数据科学家，能够生成高质量的合成数据 Autodata: An agentic data scientist to create high quality synthetic data

▲ 7

轻视重思：多模态思维推理能做什么，不能做什么 Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do

▲ 7 💬 1

GUI与CLI：仅基于屏幕操作或依赖技能的计算机使用型智能体中的执行瓶颈 GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use Agents

▲ 7 💬 1 ★ 1

GBC：用于优化多智能体系统的梯度连接方法 GBC: Gradient-Based Connections for Optimizing Multi-Agent Systems

▲ 7 💬 1

ProMSA：用于基于知识的视觉问答的渐进式多模式搜索代理 ProMSA:Progressive Multimodal Search Agents for Knowledge-Based Visual Question Answering

▲ 7 💬 1 ★ 13

基于层次结构的优势权重分配方法，用于在线强化学习中对来自稀疏片段数据的VLAs进行微调 Hierarchical Advantage Weighting for Online RL Fine-Tuning of VLAs from Sparse Episode Outcomes

▲ 6 💬 1

MVEB：大规模视频嵌入基准测试 MVEB: Massive Video Embedding Benchmark

▲ 6 💬 1

不稳定的特性、可复现的子空间：理解稀疏自编码器中种子依赖性的问题 Unstable Features, Reproducible Subspaces: Understanding Seed Dependence in Sparse Autoencoders

▲ 6 💬 1

Looped World Models

▲ 6 FaceMind Research Asia · 人工智能、机器学习、世界模型、强化学习、序列建模

ChLogic：评估中文表达中逻辑推理的稳健性 ChLogic: Evaluating Robustness of Logical Reasoning in Chinese Expressions

▲ 6 💬 2

一种基于梯度优化的RLVR稳定性分析与胜者优势策略优化方法 A Gradient Perspective on RLVR Stability and Winner Advantage Policy Optimization

▲ 6 💬 1 ★ 1

FAPO：多步骤LLM流程的完全自主优化机制 FAPO: Fully Autonomous Prompt Optimization of Multi-Step LLM Pipelines

▲ 6 💬 1 ★ 13

针对代理式和多模态LLM的上下文感知强化学习 Context-Aware RL for Agentic and Multimodal LLMs

▲ 6 💬 3 ★ 1

多轮反射掩模技术能够激发掩模扩散模型中的推理能力 Multi-Turn Reflective Masking Elicits Reasoning in Mask Diffusion Models

▲ 6 💬 1

智能体时代中的因果发现 Causal Discovery in the Era of Agents

▲ 6 💬 1 ★ 1636

迈向无风险的开放权重模型：在LLM中区分公共与私有能力 Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMs

▲ 6 💬 1 ★ 1

TryOnCrafter：通过可渲染的4D试穿代理，释放相机运动轨迹，实现逼真的视频虚拟试穿体验 TryOnCrafter: Unleashing Camera Trajectories for Realistic Video Virtual Try-on via a Renderable 4D Try-on Proxy

▲ 6 💬 1

推进基于WordArt的场景文本识别研究：数据集与方法 Advancing WordArt-Oriented Scene Text Recognition: Datasets and Methods

▲ 6 💬 1 ★ 3

用于机器人控制的上下文环境建模 In-Context World Modeling for Robotic Control

▲ 6 💬 1

用于可靠视频理解的信任感知工具调度机制 Confidence-Aware Tool Orchestration for Robust Video Understanding

▲ 6 💬 1 ★ 2

训练后被忽视的免费午餐：对LLM智能体来说是一种优势 Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

▲ 6 💬 1 ★ 2

SimFoundry：用于政策学习与评估的模块化与自动化场景生成 SimFoundry: Modular and Automated Scene Generation for Policy Learning and Evaluation

▲ 6

Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification

▲ 5 复旦大学可信具身智能研究院、上海创新研究院、阿里巴巴通义千问团队 · 计算机视觉、多模态大模型、视觉生成、视觉理解、自回归模型

Dr-DCI：通过动态工作空间扩展来实现直接语料库交互的扩展 Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion

▲ 5 💬 1 ★ 2

奖励其实一直都在你的数据中：通过判别器引导的强化学习来修正流匹配问题 The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL

▲ 5 💬 1

LedgerAgent：用于政策遵循型工具调用代理的结构化状态 LedgerAgent: Structured State for Policy-Adherent Tool-Calling Agents

▲ 5 💬 1

HumanScale：以自我为中心的人体视频在实体预训练中能够优于真实机器人数据 HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining

▲ 5 💬 2

理解具有环境感知能力的信息检索行为 Understanding the Behaviors of Environment-aware Information Retrieval

▲ 5 💬 2 ★ 5

基于视觉的思考 Thinking with Visual Grounding

▲ 5 💬 3 ★ 1

锥形语言模型 Tapered Language Models

▲ 5

文本到图像模型是归纳式“火鸡”吗？一种用于因果推理的反事实基准测试 Are Text-to-Image Models Inductivist Turkeys? A Counterfactual Benchmark for Causal Reasoning

▲ 5 💬 1

ReNIO：对LLM在线策略蒸馏中的负面轨迹重要性进行重新加权 ReNIO: Reweighting Negative Trajectory Importance for LLM On-Policy Distillation

▲ 5 💬 1 ★ 1

RL-Index：用于检索索引推理的强化学习 RL-Index: Reinforcement Learning for Retrieval Index Reasoning

▲ 5 💬 1

可变宽度变形器 Variable-Width Transformers

▲ 4 ★ 3

超越单一语言深度研究：使用跨语言检索功能来评估代理程序和检索器 Beyond Monolingual Deep Research: Evaluating Agents and Retrievers with Cross-Lingual BrowseComp-Plus

▲ 4 💬 1

PAIWorld：一种适用于机器人操作的、符合3D标准的世界模型 PAIWorld: A 3D-Consistent World Foundation Model for Robotic Manipulation

▲ 4

SciOrch：学习如何协调专业的LLM来解决前沿的多模态科学推理任务 SciOrch: Learning to Orchestrate Expert LLMs for Solving Frontier Multimodal Scientific Reasoning Tasks

▲ 4 💬 1 ★ 1

ViT-Up：用于视觉变换器的忠实特征上采样方法 ViT-Up: Faithful Feature Upsampling for Vision Transformers

▲ 4 💬 1 ★ 6

Holo-World：用于视频世界模型的统一相机、物体和天气控制系统 Holo-World: Unified Camera, Object and Weather Control for Video World Model

▲ 4 ★ 3

利用全局地图和局部视角进行多视图3D推理的丰厚回报 Dense Reward for Multi-View 3D Reasoning with Global Maps and Local Views

▲ 4 💬 1 ★ 4

PolicyTrim：提升视觉-语言-动作模型的内在政策效率 PolicyTrim: Boosting Intrinsic Policy Efficiency of Vision-Language-Action Models

▲ 4 💬 1 ★ 2

顾问：用于代理任务的多维度评估数据集 Counsel: A Meta-Evaluation Dataset for Agentic Tasks

▲ 4 💬 1

DiffusionBench：关于扩散Transformer的全面评估 DiffusionBench: On Holistic Evaluation of Diffusion Transformers

▲ 4 ★ 54

CAVEWOMAN：大型语言模型在语言输入和输出压缩条件下的行为如何 CAVEWOMAN: How Large Language Models Behave Under Linguistic Input and Output Compression

▲ 4 💬 1 ★ 1

针对KV-Cache量化的RoPE感知位分配 RoPE-Aware Bit Allocation for KV-Cache Quantization

▲ 4 💬 2 ★ 3

当较低特权就足够时：探究大语言模型智能体中的过度特权工具选择问题 When Lower Privileges Suffice: Investigating Over-Privileged Tool Selection in LLM Agents

▲ 4 💬 1 ★ 3

CoffeeBench：在异质多智能体经济环境中对长期视角的LLM智能体进行基准测试 CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies

▲ 4 💬 1 ★ 6

像素空间自回归图像生成的并行部署近似方法 Parallel Rollout Approximation for Pixel-Space Autoregressive Image Generation

▲ 4 💬 1 ★ 4

学习折叠技巧：在2026年LeHome挑战赛中成为获奖解决方案（线上第一名，线下第二名） Learning to Fold: prizewinning solution at LeHome Challenge 2026 (1st place online, 2nd offline)

▲ 4 💬 1 ★ 24

Ko-WideSearch：一种用于网络代理进行全面集合枚举的韩国式广度搜索基准测试 Ko-WideSearch: A Korean Breadth-Search Benchmark for Exhaustive Set Enumeration by Web Agents

▲ 4 💬 1

说话时的思考：用于响应型与智能对话语音代理的推理时刻知识传递 Thinking While Speaking: Inference-Time Knowledge Transfer for Responsive and Intelligent Conversational Voice Agents

▲ 4 💬 2 ★ 5

MMDiff：扩展扩散变换器以用于多模态生成 MMDiff: Extending Diffusion Transformers for Multi-Modal Generation

▲ 3 💬 1

你不需要过于复杂的假设：通过时间差异来实现视觉表示学习 You Don't Need Strong Assumptions: Visual Representation Learning via Temporal Differences

▲ 3 💬 2 ★ 3

Track2View：通过配对的三维点轨迹技术，实现与4D一致性的摄像机控制视频生成 Track2View: 4D-Consistent Camera-Controlled Video Generation via Paired 3D Point Tracks

▲ 3 💬 1

提示级蒸馏：一种非参数化的替代方法，用于实现高效的推理能力提升 Prompt-Level Distillation: A Non-Parametric Alternative to Model Fine-Tuning for Efficient Reasoning

▲ 3 💬 1

2026年人工智能指数报告 Artificial Intelligence Index Report 2026

▲ 3

LoopCoder-v2: Only Loop Once for Efficient Test-Time Computation Scaling

▲ 3 北京航空航天大学、IQuest Research、Langboat、Renming University of China · 大语言模型、代码生成、推理优化、测试时计算

视觉搜索器：通过主动视觉推理实现视觉与自然语言多模态智能搜索 Visual-Seeker: Towards Visual-Native Multimodal Agentic Search via Active Visual Reasoning

▲ 3 💬 1

CEO-评估：智能体能走长期路线吗？ CEO-Bench: Can Agents Play the Long Game?

▲ 3 ★ 3

FID抽奖：量化生成模型评估中的隐藏随机性 The FID Lottery: Quantifying Hidden Randomness in Generative-Model Evaluation

▲ 3 💬 1

JAMER：专业游戏引擎层面的代码框架数据集与评估标准 JAMER: Project-Level Code Framework Dataset and Benchmark on Professional Game Engines

▲ 3

用LOCUS释放法律：美国的地方法规体系 Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States

▲ 3 💬 1

对分辨率具有鲁棒性的自适应体积力学性质场 Adaptive Volumetric Mechanical Property Fields Invariant to Resolution

▲ 3 💬 1

LegalHalluLens：基于类型化幻觉检测与校准的多智能体辩论技术，用于实现可信的法律人工智能 LegalHalluLens: Typed Hallucination Auditing and Calibrated Multi-Agent Debate for Trustworthy Legal AI

▲ 3 💬 2 ★ 1

Taylor-Calibrate：混合线性注意力蒸馏的基于原则的初始化方法 Taylor-Calibrate: Principled Initialization for Hybrid Linear Attention Distillation

▲ 3 💬 5 ★ 1

以视频对象为中心的学习中的选择性协同学习 Selective Synergistic Learning for Video Object-Centric Learning

▲ 3 💬 2 ★ 4

WorldLines：长期展望的有状态实体化智能体的基准测试与建模 WorldLines: Benchmarking and Modeling Long-Horizon Stateful Embodied Agents

▲ 3 💬 1

GeneralVLA-2：基于几何信息的重建与受控记忆技术，用于机器人规划 GeneralVLA-2: Geometry-Aware Reconstruction and Governed Memory for Robot Planning

▲ 3 💬 1 ★ 4

FastMix：通过梯度下降实现快速数据混合优化 FastMix: Fast Data Mixture Optimization via Gradient Descent

▲ 3 💬 1 ★ 2

中间层所知道的内容：从熵动力学中检测出“ jailbreak”现象 What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics

▲ 3 💬 1 ★ 1

PrivacyAlign：用于LLM智能体的情境化隐私对齐技术 PrivacyAlign: Contextual Privacy Alignment for LLM Agents

▲ 3 💬 1

离散化奖励模型 Discretizing Reward Models

▲ 3 💬 1

NormGuard：流匹配强化学习中的保奖励范数约束 NormGuard: Reward-Preserving Norm Constraints in Flow-Matching Reinforcement Learning

▲ 3 💬 1

利用 Google Paper Assistant 工具实现科学论文的自动化评审 Towards Automating Scientific Review with Google's Paper Assistant Tool

▲ 3

以对象为中心的残差RL技术用于零样本真实场景VLA增强 Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA Enhancement

▲ 3 💬 1

PermaVid：通过解耦的上下文记忆实现跨编辑过程的一致视频生成 PermaVid: Consistent Video Generation Across Edits via Disentangled Context Memory

▲ 2 💬 1 ★ 8

RepSelect：通过表示选择性实现稳健的LLM“去学习”功能 RepSelect: Robust LLM Unlearning via Representation Selectivity

▲ 2 💬 1 ★ 1

将量子算符与大型语言模型进行对齐 Aligning Quantum Operators with Large Language Models

▲ 2 💬 2

RODS：以奖励为驱动的在线数据整合技术，用于多轮工具使用型智能体 RODS: Reward-Driven Online Data Synthesis for Multi-Turn Tool-Use Agents

▲ 2

使用图灵奖励学习用户模拟器 Learning User Simulators with Turing Rewards

▲ 2 💬 1 ★ 1

通过研究平台将人工智能科学家的研究整合、综合与验证工作外包出去 Externalizing Research Synthesis and Validation in AI Scientists through a Research Harness

▲ 2 ★ 13

MyPCBench：一种用于个人智能计算机使用智能代理的测试工具 MyPCBench: A Benchmark for Personally Intelligent Computer-Use Agents

▲ 2 💬 2 ★ 3

IndustryBench-MIPU：工业产品多图像属性值提取的基准测试 IndustryBench-MIPU: Benchmarking Multi-Image Attribute Value Extraction for Industrial Products

▲ 2 💬 1 ★ 8

iOSWorld：个人智能电话代理的评估标准 iOSWorld: A Benchmark for Personally Intelligent Phone Agents

▲ 2 💬 2

重新思考LLM FP4预训练中的压缩偏见：几何起源、系统性影响与UFP4方法 Rethinking Shrinkage Bias in LLM FP4 Pretraining: Geometric Origin, Systemic Impact, and UFP4 Recipe

▲ 2 💬 1

工作负载波动下ASR服务的时间感知调度 Duration Aware Scheduling for ASR Serving Under Workload Drift

▲ 2 💬 3

SproutRAG：基于渐进嵌入的注意力引导树搜索技术，适用于长文档RAG场景 SproutRAG: Attention-Guided Tree Search with Progressive Embeddings for Long-Document RAG

▲ 2 💬 1

MCompassRAG：主题元数据作为段落级检索的语义指南 MCompassRAG: Topic Metadata as a Semantic Compass for Paragraph-Level Retrieval

▲ 2 💬 1

SpatialAvatar-0：高质量4D头部虚拟形象，采用多阶段重建技术 SpatialAvatar-0: High-Quality 4D Head Avatar with Multi-Stage Reconstruction

▲ 2 💬 1

将示例转化为任务指令：提升现实世界B2B对话中的上下文内学习能力 Distilling Examples into Task Instructions: Enhanced In-Context Learning for Real-World B2B Conversations

▲ 2 💬 1

Vera：一种用于保留内容信息的视频编辑的多层扩散模型 Vera: A Layered Diffusion Model for Content-Preserving Video Editing

▲ 2 💬 1

UniverSat：适用于地球观测的、不依赖分辨率和模态的Transformer模型 UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation

▲ 2 💬 1 ★ 20

CalVerT：通过使用经过校准的验证器遥测数据来增强智能体，能够提升对知识密集型任务中的操作能力与学习效果 CalVerT: Augmenting Agents with Calibrated Verifier Telemetry Improves Action and Learning in Knowledge-Intensive Tasks

▲ 2 💬 1

流形强盗：基于大语言模型潜在几何特性的贝叶斯课程学习 Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models

▲ 2 💬 2

基于多目标强化学习的LLM预训练整体数据调度器 Holistic Data Scheduler for LLM Pre-training via Multi-Objective Reinforcement Learning

▲ 2 💬 1 ★ 200

QG-MIL：一种用于医学影像领域无关多实例学习的门控Transformer聚合器 QG-MIL: A Gated Transformer Aggregator for Domain-Agnostic Multiple Instance Learning in Medical Imaging

▲ 2 💬 1 ★ 1

基于信息的KV缓存压缩技术，用于长推理任务 Information-Aware KV Cache Compression for Long Reasoning

▲ 2 💬 1

PhysiFormer：学习在世界空间中模拟力学原理 PhysiFormer: Learning to Simulate Mechanics in World Space

▲ 2

ABACUS：将统一基础模型进行适配，以实现图像数量理解与生成之间的衔接 ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and Generation

▲ 2 💬 1

集群、路径、升级：基于成本意识的LLM服务级级架构 Cluster, Route, Escalate: Cascaded Framework for Cost-Aware LLM Serving

▲ 2 💬 1

AgentOdyssey：一种用于测试阶段持续学习智能体、具有开放性的长期目标文本生成游戏机制 AgentOdyssey: Open-Ended Long-Horizon Text Game Generation for Test-Time Continual Learning Agents

▲ 2 💬 1

ProCUA-SFT Technical Report

▲ 1 NVIDIA、University of Washington · 机器学习、计算机视觉、多模态大模型、智能体、桌面自动化

探索性强化学习：用于大型语言模型训练过程中的中期优化 ExpRL: Exploratory RL for LLM Mid-Training

▲ 1 💬 1 ★ 2

EgoPhys：从以自我为中心的视频中学习可泛化的可变形物体物理模型 EgoPhys: Learning Generalizable Physics Models of Deformable Objects from Egocentric Video

▲ 1 💬 1

基于大型语言模型的生成式推荐中的隐性推理机制 Implicit Reasoning for Large Language Model-based Generative Recommendation

▲ 1 💬 1

LaWAM：用于高效动态感知机器人策略的潜在世界动作模型 LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies

▲ 1 💬 1 ★ 14

谁在“翻转”？自我反驳与跨模型反驳共同揭示了大型语言模型中决策的不稳定性 Who Flips? Self- and Cross-Model Counterarguments Reveal Answer Instability in LLMs

▲ 1 💬 1 ★ 1

在噪声感知下的选择性控制：模块化网络中被综合指标隐藏的治理失败 Selective Control under Noisy Perception: Governance Failures Hidden by Aggregate Metrics in Modular Networks

▲ 1 💬 1

RefGC-SR^2：基于参考信息的生成内容超分辨率处理与优化技术 RefGC-SR^2: Reference-guided Generated Content Super-Resolution and Refinement

▲ 1 💬 1

MotionVLA：一种用于仿人机器人运动的视觉-语言-动作模型 MotionVLA: Vision-Language-Action Model for Humanoid Motion

▲ 1 💬 1 ★ 6

HiLo-Token：一种适应输入数据的高低频令牌压缩技术，用于高效的图像编辑 HiLo-Token: Input-Adaptive High-Low Frequency Token Compression for Efficient Image Editing

▲ 1 💬 1

Bag of Dims：通过维度级符号模式实现无需训练的机制可解释性 Bag of Dims: Training-Free Mechanistic Interpretability via Dimension-Level Sign Patterns

▲ 1 💬 1

物理智商验证 Physics-IQ Verified

▲ 1 💬 1

在推理之前进行感知：分离感知与推理，以实现对捷径具有抵抗力的多模态在线策略自蒸馏 Seeing Before Reasoning: Decoupling Perception and Reasoning for Shortcut-Resilient Multimodal On-Policy Self-Distillation

▲ 1 💬 2 ★ 2

Morpheus：一种了解词形的神经分词器和土耳其语词嵌入器 Morpheus: A Morphology-Aware Neural Tokenizer and Word Embedder for Turkish

▲ 1 💬 1 ★ 2

LooseControlVideo：利用空间分割技术进行导演级视频控制 LooseControlVideo: Directorial Video Control using Spatial Blocking

▲ 1 💬 1

显微镜下的数据流形 The Data Manifold under the Microscope

▲ 1 💬 4

风格偏见：一些人类视觉线索导致了MLLMs中的大多数社会偏见 StylisticBias: A Few Human Visual Cues Drive Most Social Biases in MLLMs

▲ 1 💬 1 ★ 1

当代理过早承诺时：如何识别大语言模型代理的过早承诺问题 When Agents Commit Too Soon: Diagnosing Premature Commitment in LLM Agents

▲ 1 💬 1

Arbor：用于可控3D资源生成的显式几何约束方法 Arbor: Explicit Geometric Conditioning for Controllable 3D Asset Generation

▲ 1 💬 1 ★ 2

MeshFlow：基于等变流匹配的网格生成 MeshFlow: Mesh Generation with Equivariant Flow Matching

▲ 1 💬 1 ★ 27

剧本：让LLM智能体拥有音乐结构的感知 Libretto: Giving LLM Agents a Sense of Musical Structure

▲ 1 💬 1 ★ 1

可验证的搜索并非一种可学习的思维过程 A Verifiable Search Is Not a Learnable Chain-of-Thought

▲ 1 💬 1 ★ 1

揭开数据受限语言模型预训练中训练时间增强的神秘面纱 Demystifying Training-Time Augmentation for Data-Constrained Language Model Pretraining

▲ 1 💬 1 ★ 3

Robusto-2：在利马和纽约市对自动驾驶领域的人类与虚拟智能体进行基准测试 Robusto-2: Benchmarking Humans & VLMs for Autonomous Driving in Lima & New York City

▲ 1 💬 1

AC-ODM：演员–批评家在线数据混合技术，用于高效处理样本的LLM预训练 AC-ODM: Actor--Critic Online Data Mixing for Sample-Efficient LLM Pretraining

▲ 1 💬 1 ★ 1

FlowR2A：用于多模态驾驶规划的学习奖励-动作分布方法 FlowR2A: Learning Reward-to-Action Distribution for Multimodal Driving Planning

▲ 1 💬 1 ★ 2

DREAM：通过自回归建模实现密集检索嵌入 DREAM: Dense Retrieval Embeddings via Autoregressive Modeling

▲ 1 💬 1 ★ 3

ReMMD：用于多模态虚假信息检测的现实化多语言多图像智能验证系统 ReMMD: Realistic Multilingual Multi-Image Agentic Verification for Multimodal Misinformation Detection

▲ 1 💬 1

机器人操作的世界价值模型 World Value Models for Robotic Manipulation

▲ 1

ChartWalker：对跨图表RAG任务进行基准测试 ChartWalker: Benchmarking the Cross-Chart RAG Task

▲ 1 💬 1

EventVLA：基于事件驱动的视觉证据记忆机制，用于长距离视觉-语言-动作策略 EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies

▲ 1 💬 1 ★ 14

一次蒸馏，终生适应：探索用于持续测试时适应的数据集蒸馏方法 Distill Once, Adapt Life-Long: Exploring Dataset Distillation for Continual Test-Time Adaptation

▲ 1 💬 1 ★ 2

EO-WM：一种基于物理信息的地球观测预测世界模型 EO-WM: A Physically Informed World Model for Probabilistic Earth Observation Forecasting

▲ 1 💬 1 ★ 3

世界模型中的幻觉是可预测且可预防的 Hallucination in World Models is Predictable and Preventable

▲ 1 💬 1 ★ 20

语言模型结合何时能发挥作用？在67种前沿模型中，路由、投票和混合智能体任务上的共同失败上限问题 When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models

▲ 1 💬 1

COrigami：一种用于协同设计可平铺、可识别的折纸的AI流程工具 COrigami: An AI Pipeline for Co-Designing Flat-Foldable Visually Recognisable Origami

▲ 1

OpenBioRQ：尚未解决的生物医学研究问题与药物相关 OpenBioRQ: Unsolved Biomedical Research Questions for Agents

▲ 1 💬 1 ★ 1

Qwen-RobotManip技术报告：对齐机制为机器人操作基础模型的实现提供了更大的规模支持 Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models

▲ 1 💬 1

MemoBench：在动态变化的环境中对全球模型进行基准测试 MemoBench: Benchmarking World Modeling in Dynamically Changing Environments

▲ 1 💬 1 ★ 2

运行还是不运行：分析基于LLM的程序修复中代码执行的成本效益 To Run or Not to Run: Analyzing the Cost-Effectiveness of Code Execution in LLM-Based Program Repair

▲ 1 💬 1

《Galaxy的分词器指南：科学基础模型的一个基准》 The Galaxy's Guide to the Tokenizer: A Benchmark for Scientific Foundation Models

▲ 1 💬 1 ★ 48

Qwen-RobotNav技术报告：一种适用于代理式导航系统的可扩展导航模型 Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System

▲ 1 💬 1

人类共有的感知能力 Human Universal Grasping

▲ 0 💬 2

Polymarket的幽灵：当链下交易与链上交易相遇时，会发生什么？ The Ghosts of Polymarket: When Off-Chain Matches Meet On-Chain Reverts

▲ 0 💬 1

TuneJury：一种用于改善音乐生成偏好对齐的开放式评估方法 TuneJury: An Open Metric for Improving Music Generation Preference Alignment

▲ 0 💬 1 ★ 9

对机器文本检测器的攻击仍然保留着特定的风格特征 Attacks on Machine-Text Detectors Retain Stylistic Fingerprints

▲ 0 💬 1 ★ 2

验证性检测与预防多智能体大型语言模型系统中的并发异常行为 Verified Detection and Prevention of Concurrency Anomalies in Multi-Agent Large Language Model Systems

▲ 0 💬 1

无政府状态下的分散推理代价 The Price of Anarchy in Disaggregated Inference

▲ 0 💬 1

用于评估电子表格中下一步行动预测的基准与框架 A Benchmark and Framework for Evaluating Next Action Predictions in Spreadsheets

▲ 0 💬 1

基于LLM的NWDAF：迈向AI原生型6G网络智能化的关键一步 LLM-Enabled NWDAF: A Step Toward AI-Native 6G Network Intelligence

▲ 0 💬 1

可配置的临床信息提取与代理式RAG技术：哪些方法有效，哪些会失败，以及原因是什么 Configurable Clinical Information Extraction with Agentic RAG: What Works, What Breaks, and Why

▲ 0 💬 1

没有资源，没有参照标准，就没有问题吗？在无资源语言中评估与改进用于代码生成的LLM No Resource, No Benchmarks, No Problem? Evaluating and Improving LLMs for Code Generation in No-Resource Languages

▲ 0 💬 3 ★ 1

ReSyn：一种通用的递归正则表达式合成框架 ReSyn: A Generalized Recursive Regular Expression Synthesis Framework

▲ 0 💬 4

有能力但粗心：使用计算机的智能体是否遵循上下文完整性？ Capable but Careless: Do Computer-Use Agents Follow Contextual Integrity?

▲ 0 💬 1 ★ 1

HAKARI-Bench：一种轻量级基准测试工具，用于在统一条件下比较检索架构与效率设置 HAKARI-Bench: A Lightweight Benchmark for Comparing Retrieval Architectures and Efficiency Settings under Unified Conditions

▲ 0 💬 2 ★ 3

关于无限跑游戏中LLM辅助重构与游戏玩法功能生成的探索性案例研究 An Exploratory Case Study of LLM-Assisted Refactoring and Gameplay Feature Generation in an Endless Runner Game

▲ 0 💬 1 ★ 1

利用人类偏好奖励提升文本转音乐生成能力 Improving Text-to-Music Generation with Human Preference Rewards

▲ 0 💬 1 ★ 1

跟随轨迹：基于点跟踪的视频合成与运动控制 Go-with-the-Track: Video Compositing and Motion Control with Point Tracking

▲ 0 💬 1 ★ 3

朝向停车位占用识别：一种自监督方法 Toward Parking Spot Occupancy Recognition: A Self-Supervised Approach

▲ 0 💬 1 ★ 2

AGORA：一种基于档案的代理式工作场景文档推理基准 AGORA: An Archive-Grounded Benchmark for Agentic Workplace Document Reasoning

▲ 0 💬 1

FLUX3D：具有扩散对齐的稀疏表示的高保真3D高斯分布生成技术 FLUX3D: High-Fidelity 3D Gaussian Generation with Diffusion-Aligned Sparse Representation

▲ 0

Multi4D：通过多层竞争分配实现高保真动态高斯撒布技术 Multi4D: High-Fidelity Dynamic Gaussian Splatting via Multi-Level Competitive Allocation

▲ 0 💬 1

一种利用局部优化方法在不确定性条件下对微电网进行最优控制的高效方法 An Efficient Method for the Optimal Control of Microgrids Under Uncertainties using Local Reduction

▲ 0 💬 2 ★ 1

开放权重LLM中的约束税：在结构化输出约束下工具调用抑制的实证研究 Constraint Tax in Open-Weight LLMs: An Empirical Study of Tool Calling Suppression Under Structured Output Constraints

▲ 0 💬 2

Lite Any Stereo V2：更快、更强大、更高效的零样本立体匹配技术 Lite Any Stereo V2: Faster and Stronger Efficient Zero-Shot Stereo Matching

▲ 0 💬 1 ★ 99

培训后的过程如何塑造生物推理模型 How Post-Training Shapes Biological Reasoning Models

▲ 0 💬 1

代码智能体需要多少静态结构？确定性锚定研究 How Much Static Structure Do Code Agents Need? A Study of Deterministic Anchoring

▲ 0 💬 1

低通道EEG智能体的边界感知上下文锚定机制 Boundary-Aware Context Grounding for A Low-Channel EEG Agent

▲ 0 💬 1 ★ 3