[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-embodiedskills-vla-verify-loop":3,"topics-all":38,"news-related-051084dc-3da0-451e-9c6b-a267d5b0e77f":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"051084dc-3da0-451e-9c6b-a267d5b0e77f","给机器人技能装上门禁:EmbodiedSkills 预检+验证闭环,RoboTwin 50 任务冲到 86.2%","浙大团队提出 EmbodiedSkills 框架:把机器人每次技能决策当\"执行提案\",运行时先查前置条件、执行后验证结果,Qwen3-VL 规划加 pi0.5 执行,RoboTwin 2.0 上 50 任务平均成功率 86.2%,LIBERO 达 97.4%。消融显示去掉中间验证成功率掉到 48.2%。","机器人执行长任务时,最大的坑不是\"不会做\",而是\"做了但不知道自己做得对不对\"。视觉-语言-动作(VLA)模型能把画面和指令直接映射成机器人动作,但一个动作预测本身并不保证它在当前物理状态下合法,也不保证执行结果会被检查。9 月 1 日上传 arXiv 的论文(2609.01281)给出了一个工程答案:EmbodiedSkills,一个把编排、训练、部署统一进单一 agent 循环的 VLA 框架。\n\n## 核心设计:技能决策变成\"执行提案\"\n\n这个框架最有意思的设定,是它对待模型决策的态度:每一次技能决策都只是一个\"提案\"。运行时在执行前检查前置条件——所需的观测、计划、后端是否就绪;执行后再用新鲜观测做结果验证。整个过程被组织成六阶段循环:观察、规划、预检、执行、验证、恢复。一条共享的可执行技能接口把高层技能选择、有边界的低层 VLA 执行和动作后验证串在同一个循环里。\n\n因为这个接口保持固定,低层 VLA 策略可以随时替换或适配,不用改动 agent 循环本身。这套接口还会把规划、执行、验证、恢复事件记录成结构化轨迹——它们既是各组件的训练监督信号,也能在有交互反馈的场景下支持可选的在线适应。论文配了 17 位作者,已开源在 GitHub(DCDmllm\u002FEmbodiedSkills)。\n\n## 数字说话:验证环节值 38 个百分点\n\n实验基于 Qwen3-VL 与 OpenPI\u002Fpi0.5,在 RoboTwin 2.0 与 LIBERO 上验证。任务适配后的低层 VLA 策略在 RoboTwin 2.0 的 50 个任务上拿到 86.20% 平均成功率(每任务 100 episode,参考基线 pi0.5 为 82.74%),LIBERO 四个套件平均 97.40%(OpenPI 参考 96.85%)。\n\n更关键的是消融:同样 50 任务、5000 episode 的受控对比里,完整 AgentLoop 是 86.20%;去掉中间验证,掉到 48.2%;用完整任务指令替代语义子任务,34.4%;把每个子任务限制为单次动作块,只剩 19.5%。也就是说,\"执行后检查一下\"这一步,单独值 38 个百分点。框架在四个依赖记忆的 RMBench 任务上平均只有 12.5%,论文没有回避:记忆仍是短板。\n\n## 所以呢\n\n这份工作把具身智能从\"单步预测准不准\"拉回到\"系统闭环靠不靠谱\"的问题上。86.2% 与 48.2% 之间的差距说明,对物理世界的长程任务,验证与恢复不是锦上添花,而是承重墙。对做机器人 agent 的团队,这份开源框架的可检查、可训练的 agent 层,是一个值得拆开看的工程参考——先把\"做了没有\"变成\"做对了没有\",再谈泛化。\n\n论文与代码:https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.01281","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.01281","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"0b9989ff-f8b0-478e-86ba-b659ec970680","en","EmbodiedSkills gives VLA robots a verify loop: 86.2% on RoboTwin","EmbodiedSkills checks each skill before execution and verifies it after. It hits 86.2% across 50 RoboTwin 2.0 tasks and 97.4% on LIBERO.","When robots fail at long-horizon tasks, the bottleneck is usually not \"can't do the action\" but \"did it without knowing whether it worked.\" Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, yet an action prediction by itself does not guarantee that the proposed operation is valid in the current physical state, or that its outcome will ever be checked. A paper uploaded to arXiv on Sep 1 (2609.01281) offers an engineering answer: EmbodiedSkills, a unified framework that orchestrates, trains, and deploys VLA agents within a single agent loop.\n\n## Core design: every skill decision is an execution proposal\n\nThe most interesting choice here is how the framework treats model decisions: every skill decision is merely a proposal. The runtime checks its prerequisites before execution — whether the required observations, plan, and backend are ready — and verifies the outcome afterward using fresh observations. The whole process runs as a six-stage loop: observation, planning, preflight, bounded execution, verification, and recovery. A shared executable-skill interface connects high-level skill selection, bounded low-level VLA execution, and post-action verification within one loop.\n\nBecause this interface stays fixed, low-level VLA policies can be replaced or adapted without touching the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories — supervision signals for individual components, and a basis for optional online adaptation when interactive feedback is available. The paper lists 17 authors and the code is open-sourced on GitHub (DCDmllm\u002FEmbodiedSkills).\n\n## The numbers: verification alone is worth 38 points\n\nThe experiments instantiate the framework with Qwen3-VL and OpenPI\u002Fpi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies reach an 86.20% average success rate across 50 RoboTwin 2.0 tasks (100 episodes per task; the pi0.5 reference baseline is 82.74%) and 97.40% across the four LIBERO suites (OpenPI reference: 96.85%).\n\nThe ablations matter more: in a controlled comparison over the same 50 tasks and 5,000 episodes, the full AgentLoop reaches 86.20%; removing intermediate verification drops it to 48.2%; replacing semantic subtasks with the full task instruction yields 34.4%; limiting every subtask to a single action chunk leaves just 19.5%. In other words, the \"check after execution\" step alone is worth 38 percentage points. On four memory-dependent RMBench tasks, the same task-adapted approach averages only 12.5% — the paper does not hide it: memory remains the weak spot.\n\n## So what\n\nThis work pulls embodied AI back from \"is the single-step prediction accurate\" to \"is the closed loop trustworthy.\" The gap between 86.2% and 48.2% shows that for long-horizon physical tasks, verification and recovery are not decoration but load-bearing walls. For teams building robot agents, the open-sourced framework's trainable and inspectable agent layer is an engineering reference worth dissecting — turn \"did it act\" into \"did it act correctly\" first, then talk about generalization.\n\nPaper and code: https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.01281","embodiedskills-vla-verify-loop","2026-09-08T17:10:00Z","2026-09-08T17:10:09.048741Z","2026-09-08T17:10:09.048749Z",true,"agent",131,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"created_at":55,"modified_at":56},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"089f56f3-32ff-4036-89b5-728d5f5a9359","边聊边干活:腾讯混元开源全模态交互 Agent Gander,小脑管对话、大脑管执行","hunyuan-gander-omni-interaction-agent","2026-09-09T21:07:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"a7095e70-a6c5-46b6-a616-446825bce5cf","Editable Visual Design：腾讯混元论文让 AI 设计稿带真实图层，直接拖拽改版","tencent-editable-visual-design-agent","2026-09-06T13:08:40+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"089195fb-7fe5-4ba9-a4bc-8e356fe5e923","BAAI把1000个GitHub仓库蒸馏成5000个技能,科研agent奖牌率31%冲到73%","baai-disco-repo-to-skill-library","2026-09-03T17:07:35+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"9a66407a-b79a-4a12-a4ef-b0d7018c8415","字节Seed新论文:VLM操作3D编辑器摆家具,把真实房间变成仿真场景","lucida-vlm-gizmoact-real-to-sim","2026-09-01T19:10:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"4c7f5330-3aff-458a-9ef5-f04cc5585703","微信视觉团队开源 WeMM 嵌入模型:2B 反超 8B 前基线,9B 达 MMEB-v2 80.6","wemm-embedding-wechat-multimodal","2026-08-26T21:07:30+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"ff0bc92a-295a-4707-be8d-76115fe9eeee","PerceptionBench 出炉:16 个前沿多模态模型,视觉感知无一及格","moonshot-perceptionbench-atomic-perception","2026-08-26T13:15:00+00:00"]