ChatPaper

AI 论文热榜

近 7 天全球 AI 研究者都在点赞的论文，人话解读

  1. 1

    Repo-To-Skill：将 GitHub 代码仓库蒸馏为 AI4AI 技能

    🔥 4942026-09-02深入阅读arXiv 原文Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills
  2. 2

    StudentSim：基于大语言模型的学生模拟器训练

    🔥 4582026-09-01深入阅读arXiv 原文StudentSim: Training LLM-based Student Simulators
  3. 3

    Qwen-Drive-1.0：迈向自动驾驶视觉语言基础模型的初步探索

    🔥 3372026-08-31深入阅读arXiv 原文Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving
  4. 4

    HarnessDev：大语言模型能否创建并演进自身的智能体框架？

    🔥 2242026-09-01深入阅读arXiv 原文HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?
  5. 5

    Terminal-Universe：将智能体轨迹转化为可扩展的终端环境

    🔥 2132026-09-03深入阅读arXiv 原文Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments
  6. 6

    LLaDA-Image：以全开放训练方案构建强大的图像生成器

    🔥 1962026-09-03深入阅读arXiv 原文LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes
  7. 7

    Aspire：模型能否从模糊目标中自我进化？

    🔥 1732026-08-31深入阅读arXiv 原文Aspire: Can Models Self-Evolve from Vague Goals?
  8. 8

    懂得何时不重用：自主LLM后训练中的条件化经验迁移

    🔥 1392026-08-27深入阅读arXiv 原文Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training
  9. 9

    SolarWM：面向长时程视频世界模型的开放数据与可扩展训练

    🔥 1332026-09-02深入阅读arXiv 原文SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models
  10. 10

    随机注意力：重新思考KV缓存清除以实现高效推理

    🔥 1232026-09-03深入阅读arXiv 原文Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
  11. 11

    EarlyEval：基于早期结果预测的低成本智能体评估

    🔥 1102026-09-02深入阅读arXiv 原文EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction
  12. 12

    LatentPress: Context Compression Beyond Text and Vision

    🔥 1022026-09-01深入阅读arXiv 原文
  13. 13

    在策略蒸馏真的在蒸馏吗？从噪声教师到自我提升

    🔥 982026-08-31深入阅读arXiv 原文Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement
  14. 14

    LoopArena：模型作为循环工程运行时控制器的基准测试

    🔥 982026-08-28深入阅读arXiv 原文LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering
  15. 15

    DreamX-Creator：普及2K分辨率的原生音视频生成

    🔥 902026-08-31深入阅读arXiv 原文DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution
  16. 16

    DART-SD：面向多轮工具调用智能体自蒸馏的菱形拓扑感知检索与调优

    🔥 882026-08-19深入阅读arXiv 原文DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents
  17. 17

    超越数据扩展：面向视觉-语言-动作模型的表征中心持续预训练

    🔥 842026-08-27深入阅读arXiv 原文Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Models
  18. 18

    SMELT: 计算匹配的MoE循环Transformer的缩放定律

    🔥 752026-09-01深入阅读arXiv 原文SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers
  19. 19

    Lucida：面向可组合真实到仿真场景建模的解析、生成与放置

    🔥 722026-08-31深入阅读arXiv 原文Lucida: Parse, Generate, and Place for Composable Real-to-Sim Scene Modeling
  20. 20

    为何门控DeltaNet能经受4比特量化考验：混合27B大语言模型中递归部分的NVFP4 W4A4量化方案

    🔥 662026-09-03深入阅读arXiv 原文Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM

每天 5 分钟，跟上 AI 前沿

每日精选 5 篇热门论文，讲成人话，配一道小谜题。

去看今日刊 →