[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-code2skill-agent-skill-synthesis":3,"topics-all":38,"news-related-b0c434a5-4911-4297-b1ef-2c44cbc26653":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},"b0c434a5-4911-4297-b1ef-2c44cbc26653","蚂蚁新研究:19769 个代码仓库,炼出百万条 agent 技能","蚂蚁国际团队发布 Code2Skill:从 19769 个 GitHub 仓库自动合成 100.6 万条可验证技能记录,检索增强后 72 组配对评测平均提升 11.7%,在 7 个共享基准上全面超过轨迹合成路线;数据集已开放,代码仓库仍是 pre-release。","编程 agent 的「技能」从哪来?过去有两条主流路:让 agent 在真实环境里摸爬滚打,把交互轨迹沉淀成技能;或者从文档里抽取操作说明。前者要付出昂贵的交互成本,后者缺乏可执行证据、真假难辨。9 月 4 日挂上 arXiv 的一篇论文(arXiv:2609.05571)给出了第三条路:直接从源代码合成技能。这项来自蚂蚁国际(Ant International)的工作,已登上 Hugging Face Daily Papers 9 月 21 日榜单的第一位,拿到 86 个赞,由论文一作亲自提交。\n\n## 方法:看不见源码的「盲态重建」\n\nCode2Skill 流水线分七步:先给解析出的源代码单元(函数、方法、命令入口、文件级组件)打分,挑出有复用价值的程序性知识;再抽取成三类带类型的技能记录——原子操作、组合流程、常见模式。核心验证环节是「盲态重建」:模型只拿到技能描述和接口,看不到源码本体、仓库名和文件路径,要凭描述把代码重新写出来;随后由独立的等价判定环节对比源码与重建代码,裁决阶段还会区分「技能本身站不住」和「重建失败」两种情况,避免误杀合格技能。全部通过才计入技能库,并带上工作流、边界、来源与证据四类元数据。\n\n## 数字:百万条记录与 72 组配对评测\n\n论文报告的规模相当可观:19769 个热门且活跃维护的 GitHub 仓库,最终产出 1006822 条合格记录。效果上,检索接入这些技能后,模型在 9 种模型设置、8 个基准共 72 组配对评测中平均提升 11.7%,57 组胜出;与轨迹合成的技能库在 7 个共享基准上对比全部占优,说明仓库挖掘的技能在 agent 还没攒够交互经验时就能用。更值得注意的一组数字:从 AI 生成代码里合成的技能,通过率 93.50%,人工写出的代码是 93.00%,几乎持平——意味着随着 AI 写的代码越来越多,这条管线可以跟着一起膨胀,不会遇到素材枯竭。\n\n## 冷水:判定不是证明,代码也还没开源\n\nGitHub 仓库的 README 写得很直白:判定环节「是 LLM 一致性检查,不是程序等价性证明」。也就是说,百万条记录的「合格」标准仍然是模型说了算,不是数学意义上的验证。另一个现实是仓库状态:目前是 pre-release,只有 3 个 commit,没有附许可证,README 明确写着「还不是一次公开的开源发布」。真正开放的是配套数据集 DeveloperSkills-Code2Skill,已放在 Hugging Face 上,论文一作也在评论区给出了链接。想复现整条流水线的人,眼下只能等。\n\n## 意义:技能获取从攒经验变成挖资产\n\n这项工作把「技能获取」从经验积累重构为资产挖掘:GitHub 上沉淀的海量代码,本身就是人类程序性知识最大的现成载体,现在有了一条可验证(尽管验证者仍是 LLM)的批量提取路径。对蚂蚁国际来说,这是 Ling 系列 Agent 模型之后在 agentic 基础设施上的又一次下注。至于 LLM 判定的可靠性天花板在哪里,恐怕要等第三方在真实 agent 工作流里复现之后才有答案——这也是所有「合成数据 + 模型自证」路线共同悬着的问题。\n\n参考:arXiv:2609.05571(arxiv.org\u002Fabs\u002F2609.05571)、GitHub ant-intl\u002FCode2Skill、数据集 huggingface.co\u002Fdatasets\u002Fant-intl\u002FDeveloperSkills-Code2Skill","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.05571","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},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":19,"name":20,"slug":20,"description":14,"color":14},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"7ef2e6e0-f269-4dff-961a-5d28e92f1c31","en","Code2Skill Turns 19,769 GitHub Repos Into a Million Agent Skills","Ant International mines 1M+ agent skills from 19,769 GitHub repos; retrieval adds 11.7% across 72 evals. Dataset open, code still pre-release.","Where do coding agent \"skills\" come from? Two routes have dominated: let agents accumulate skills through real interactions, or extract procedural knowledge from documentation. The first is expensive — every skill costs environment interaction; the second lacks executable evidence and is hard to verify. A paper published on arXiv on Sep 4 (arXiv:2609.05571) from Ant International proposes a third route: synthesize skills directly from source code. As of the Sep 21 Hugging Face Daily Papers board it ranks #1 of the day with 86 upvotes, submitted by the first author.\n\n## Method: source-body-blind regeneration\n\nThe Code2Skill pipeline runs in seven steps: it first scores parsed source units — functions, methods, command entry points, file-level components — for reusable procedural evidence; then extracts typed skill records in three classes: atomic operations, composite workflows, and recurring patterns. The core validation step is \"source-body-blind regeneration\": a model receives only the skill description and interface — not the source body, repository name, or file path — and must regenerate the code from that alone. A separate equivalence judgment compares source and regenerated code, and an adjudication stage separates \"the skill is unsupported\" from \"the regeneration failed\", so good skills are not discarded by accident. Only records passing everything enter the bank, tagged with workflow, boundary, provenance, and source-evidence metadata.\n\n## Numbers: a million records, 72 matched evaluations\n\nThe reported scale is substantial: 19,769 popular, actively maintained GitHub repositories yield 1,006,822 accepted records. With retrieved skills attached, models improve by 11.7% on average across 72 protocol-matched evaluations spanning nine model settings and eight benchmarks, winning 57 of them; against trajectory-derived skill banks, code-derived skills win on all seven shared benchmarks — meaning repository-mined skills help before an agent has accumulated interaction experience. One more striking pair: skills synthesized from AI-generated code pass at 93.50%, versus 93.00% for human-written code, nearly identical. As AI-generated code keeps growing, this pipeline can scale with it rather than run out of material.\n\n## Caveats: the judge is not a proof, and the code is not open yet\n\nThe GitHub README is blunt: the judge \"is an LLM consistency check, not a proof of program equivalence.\" The million-record \"accepted\" standard is still model-decided, not mathematically verified. The repo reality: it is a pre-release with only 3 commits and no license attached; the README states plainly that it \"is not yet a public open-source release.\" What is actually open is the companion dataset, DeveloperSkills-Code2Skill on Hugging Face — the first author shared the link in the paper's comment section. Anyone wanting to reproduce the full pipeline will have to wait.\n\n## Why it matters\n\nThis work reframes skill acquisition from accumulated experience to asset mining: the code already sitting on GitHub is the largest existing carrier of human procedural knowledge, and there is now a batch extraction path that is verifiable — albeit verified by an LLM. For Ant International, it is another agentic-infrastructure bet after the Ling series of agent models. Where the LLM-judge reliability ceiling sits, though, will only be answered when third parties reproduce it inside real agent workflows — the same open question hanging over every \"synthetic data + model self-verification\" pipeline.\n\nReferences: arXiv:2609.05571 (arxiv.org\u002Fabs\u002F2609.05571); GitHub ant-intl\u002FCode2Skill; dataset huggingface.co\u002Fdatasets\u002Fant-intl\u002FDeveloperSkills-Code2Skill","code2skill-agent-skill-synthesis","2026-09-21T19:06:31Z","2026-09-21T19:06:42.258262Z","2026-09-21T19:06:42.258277Z",true,"agent",56,[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},"b0c4e8d2-5662-4e3e-b489-6202eabbe97b","Dream-RSI 把历史当模拟器:162 倍杠杆重写 RSI 算力账本","dream-rsi-replay-simulator-162x","2026-09-16T06:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"1bf3f192-68e9-49f1-8bbb-a350f05c15e2","LLM 编排云边端:FRAME 2026 论文点出反馈层缺环","llm-cloud-continuum-frame-2609","2026-09-16T11:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"2266cea6-06f1-4932-8905-1bc3f2e5a8c0","Meta FAIR 字节蒸馏研究:End-Of-Token 渐近反超 token 蒸馏 4%,数据只需 1\u002F6","meta-fair-byte-distillation-token-ceiling-2026-09","2026-09-15T02:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"21c7dec1-f68e-4641-974c-ae2bce87393e","教师打分、验证器掌舵:腾讯混元 FlowBalance 给自蒸馏装上方向门控,Qwen3-8B 数学均值超 GRPO 2.12 分","flowbalance-verifier-gated-self-distillation","2026-09-08T15:08:17+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"00346b75-f071-42fd-ae16-db4c5569f01a","EarlyEval 提前叫停注定失败的 Agent:近半 token 省下,分辨率只动一两个点","earlyeval-early-stop-agent-eval","2026-09-03T21:04:52+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"f65e204c-0115-4b50-9113-2c3bb2ff6637","ReCache:给 Agent 的工具记忆装上独立缓存,KV 内存砍 92%、首 token 提速 3.655 倍","recache-agent-kv-cache-reuse","2026-08-24T15:30:00+00:00"]