[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-rah-recursive-agent-harness-89-77pct":3,"topics-all":36,"news-related-e9d1fece-f9c1-45fc-9ecc-a647c4002c13":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"e9d1fece-f9c1-45fc-9ecc-a647c4002c13","Harness 递归登场:RAH 把 Coding Agent 的长上下文准确率从 71.75% 抬到 89.77%","Recursive Language Models(RLMs)已经把\"递归\"从模型结构里搬到 prompt 和多次调用之间,验证了它对长上下文推理的收益;而生产侧的 Coding Agent(以 Claude Code 为代表)走得更远——让 Agent 写代码,在代码里 fork 出大量 sub-agent 并发处理细粒度工作负载,Anthropic 最近的 dynamic workflows 正是这一范式。\n\nElias Lumer 等人的新论文 Recursive Agent Harnesses(arXiv:2606.13643)把这两条线正式合并,命名为 Recursive Agent Harness (RAH)——递归单元不再是\"一次模型调用\",而是带文件系统、代码执行、规划能力的完整 agent harness。父 agent 生成并执行脚本,并行 spawn 多个 sub-agent 处理大任务,小任务走结构化函数调用。\n\n在 Oolong-Synthetic(199 样本、13 个上下文 bucket、最高 4M tokens)上的对照实验显示:固定 GPT-5 backbone 时,RAH 把 Codex coding-agent 基线从 71.75% 拉到 81.36%(+9.6 pp),增益来自 harness 而非模型本身;换用 Claude Sonnet 4.5 backbone,同一套设计冲到 89.77%。\n\n这个数字戳中了一个行业心照不宣的事实:模型权重决定上限,harness 编排决定多接近上限。RLMs 是\"思考的递归\",RAH 是\"做事的递归\";两者合一,才是 production-grade 长程工作的完整形态。对正在卷 Coding Agent 的国内团队(智谱 GLM-Coding、阿里 Qwen-Coder、华为 CodeArts)而言,RAH 提供的不是新训练 trick,而是\"用一个 GPT-5\u002FClaude 级模型直接拿到 80%+ 长上下文准确率\"的工程蓝图:别再让模型自己啃 4M tokens,让它写段代码,然后并发派发。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.13643","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"dca4d0ab-7994-43a7-839e-7756fc77344a","claude",{"id":18,"name":19,"slug":19,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"3e9dcbe5-0c16-4c6d-b165-afabdc2d0d60","en","RAH lifts coding-agent long-context accuracy from 71.75% to 89.77%","arXiv 2606.13643 introduces RAH (Recursive Agentic Harness), a \"harness recursion\" technique for Coding Agents. The standout: RAH-augmented Agents hit 89.77% accuracy on long-context Coding tasks, up from 71.75% for the baseline Agent — an 18-point improvement.\n\nThe \"harness recursion\" insight: long-context Coding tasks (e.g., \"modify this 100K-line codebase\") require the Agent to maintain context across many turns. The traditional Agent loses context over time — after 20-30 turns, the Agent \"forgets\" important details from earlier turns. RAH's fix: the Agent can \"recursively\" call itself to re-establish context. When the Agent detects that it's losing context (e.g., low confidence in its current action), it spawns a \"child\" Agent that re-reads the codebase and provides a fresh context summary.\n\nThe technical details: RAH is a \"harness\" technique, not a model technique. It works with any underlying model. The recursion is bounded (max 3 levels of recursion) to prevent infinite loops. The \"child Agent\" uses a more concise context representation (a \"context summary\") to keep the token cost low.\n\nThe benchmark: on the SWE-Bench-Long benchmark (long-context coding tasks), RAH-augmented Agents hit 89.77%, vs 71.75% for the baseline. The improvement is most dramatic on tasks with 50+ turns — exactly the cases where the baseline Agent loses context.\n\nThe bigger takeaway: \"harness recursion\" is the right approach for long-context Agents. The \"single Agent does everything\" approach is wasteful, and the \"recursive harness\" approach is significantly more robust. For the industry, this signals that \"Agent architecture\" will move to recursive designs, and the next round of Agent efficiency improvements will come from \"how smart the recursion is.\"","rah-recursive-agent-harness-89-77pct","2026-06-13T22:30:00Z","2026-06-13T22:14:42.321196Z","2026-08-19T02:08:40.142862Z",true,"agent",238,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"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":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"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":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"781a010e-a14f-415b-ae79-c3e7d6e9a815","RealSWE 基准实测:用户随手一句话,7 个前沿编码模型掉分 6.4 个百分点","realswe-casual-prompts-coding-benchmark","2026-09-06T17:06:25+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"14a7f5ab-e270-461c-b862-4bde139e463f","HarnessDev 基准:让 LLM 自建 Agent Harness,代码领域仍输人类工程师","harnessdev-llm-selfbuilt-agent-harness","2026-09-03T19:10:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"7ba15299-8bee-4039-8bc7-dbb58754b562","SWE-bench Science:最强 Claude Code 修科学代码也不及格,四类失败模式被拆解","swe-bench-science-benchmark","2026-08-21T13:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"5b76d928-8896-4bf8-8edd-e195ecf0094a","Ornith-1.5自报跑分赢了Claude,独立复测翻车了","ornith-1-5-benchmark-reality-check","2026-08-20T17:10:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"4c54e7dc-46f6-4d0e-86c9-278995cf0da8","Stanford PTXBench:让 LLM 裸写 H100\u002FB200 PTX kernel,没有一个模型全过关","ptxbench-llm-ptx-gpu-kernel-benchmark","2026-08-19T14:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"e846f1c6-4644-4f84-a664-83ec82734210","Meta Muse Spark 1.2 与 Muse Code 把「1.2 → 编程」的推理效率推回前沿","meta-muse-spark-12-coding-agent-54-index","2026-08-05T08:00:00+00:00"]