[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-rsi-anything-world-model-recursive-self-improvement":3,"topics-all":38,"news-related-5edbfca7-7c1a-4df3-8577-a0195408b827":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},"5edbfca7-7c1a-4df3-8577-a0195408b827","RSI-Anything:世界模型陪跑,自我改进有了回滚键","arXiv新框架RSI-Anything给递归自我改进加保险:反馈贵就用JEPA世界模型预测后果挑动作,预测漂移就让模型随policy适配,升级过兼容检查才保留。一套循环跑通代码、GUI、实验规划、反应优化、免疫五域,实验室分数翻倍到46.01%,已开源。","RSI(recursive self-improvement,递归自我改进)是 agent 领域的圣杯叙事,但落地一直卡在两个现实问题上:真实环境的反馈太贵,每一步都靠试错烧不起;用世界模型替代真实反馈,预测又会随着 policy 进化越来越不可靠。10 月 9 日提交到 arXiv 的 RSI-Anything(arXiv:2610.13912)给出了一个工程味很浓的解法:让世界模型跟着 policy 一起进化,并给每一轮升级加上兼容性门控。\n\n## 一个循环跑五个领域\n\n框架核心是一条六步循环:Propose → Predict → Select → Evaluate → Learn → Retain。策略或 workflow 先提出候选动作,JEPA 世界模型预测后果,系统挑出值得真执行的动作去拿外部反馈;反馈同时驱动策略改进和世界模型的条件化适配,最后由组件检查加兼容性检查决定这对 policy-model 组合是否保留到下一轮。\n\n它的底座是同组织 9 月发布的跨域预测框架 JEPA-Anything(arXiv:2609.20800),这次相当于在其上盖了第二层楼:论文把同一套循环跑通代码修复、GUI 交互、实验室规划、反应优化、免疫科学五个领域,每个领域只换 domain adapter(状态、动作空间、反馈信号),循环本身不变。\n\n## 数字怎么说\n\n代码修复的对照设计最能说明机制价值:coder 固定不动,静态 verifier 选 patch 成功率 26.3%,换成世界模型预测选择升到 29.2%,再做世界模型适配后达 31.2%,测试用量相当,增量全部来自\"预测选得准\"。扩展到 APPS Introductory 1000 题和 LiveCodeBench v6 1055 题,最终全量测试成功率分别为 46.37% 与 33.05%,骨干为 ReasonFlux-Coder-7B。\n\nGUI 侧,MiniWoB++ 的 click-checkboxes 任务上,三个保留版本的任务完成数是 51 → 79 → 99,平均动作数从 5.30 降到 3.21,重复点击从 237 次清零;OSWorld 桌面任务相对初始 agent 成功率提升 18.4%。实验室规划(LabHorizon F200)五个版本把分数从 21.80% 拉到 46.01%,翻倍有余。化学的 Suzuki 反应优化里,同样 40 次实验预算下平均最优产率 96.15%,对照贝叶斯优化为 82.16%,且所有 run 在第 10 次观察前就找到池内最优。\n\n## 值得在意与需要冷静的地方\n\n值得在意两点。其一,RSI 的瓶颈从来不是\"会不会改\",而是\"改坏了怎么办\"——这篇把 retention 做成显式机制,升级要过组件与兼容性两道检查才保留,相当于给自我改进加了回滚,这是工程上认真的信号。其二,Gen-Verse 三个月内连出 JEPA-Anything 与 RSI-Anything,预测式世界模型正从论文概念变成可复用的基础设施:仓库 402 star、Apache-2.0 许可、配套 HuggingFace 数据集和 rsi 命令行工作流,复现门槛不高。\n\n需要冷静的也直说:上述数字全部来自论文与官方仓库自报,尚无独立复现;96.15% 是 Suzuki 反应筛选池内的对照结果,离\"AI 会做化学\"还远;免疫科学应用目前是工作流演示加前瞻假说提名(SIRPα 抑制联合 IL-36γ 轴调节),不是已验证的疗法发现。\n\n对做 agent 工程的读者,这条循环值得亲手跑一遍(rsi plan 加 rsi run 两条命令就能起);对只看故事的人,记住一句就够:自我改进的下一步,不是更猛的 policy,而是一个知道自己什么时候不可靠的世界模型。\n\n论文:https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.13912\n代码:https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FRSI-Anything","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.13912","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},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"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},"f3c1a0ee-98b1-4e5e-8735-2a485fc034dc","en","RSI-Anything: A World Model That Evolves With the Policy","JEPA world models drive recursive self-improvement in RSI-Anything: predict, select, adapt, gate retention. Lab scores double to 46.01%, code open.","Recursive self-improvement (RSI) is the holy-grail narrative of AI agents, but it has always stumbled on two practical problems: real-world feedback is expensive, and world-model predictions drift out of reliability as the policy evolves. RSI-Anything, submitted to arXiv on Oct 9 (arXiv:2610.13912), offers an engineering-flavored answer: let the world model evolve together with the policy, and gate every upgrade with compatibility checks.\n\n## One loop, five domains\n\nThe core is a six-step cycle: Propose → Predict → Select → Evaluate → Learn → Retain. A policy or workflow proposes candidate actions; a JEPA world model predicts their consequences; the system picks which actions deserve real execution and external feedback. That feedback drives both policy improvement and conditional world-model adaptation, while component and compatibility checks decide whether the updated policy-model pair is retained for the next round.\n\nThe foundation is JEPA-Anything (arXiv:2609.20800), the same group's cross-domain predictive framework released in September — this paper is effectively a second floor on top of it. The same loop runs code repair, GUI interaction, laboratory planning, reaction optimization and immune science, swapping only a domain adapter per field.\n\n## What the numbers say\n\nCode repair shows the mechanism best: with the coder fixed, a static verifier picks patches at 26.3% success; predictive selection lifts it to 29.2%, and world-model adaptation to 31.2% at comparable test usage. Extended to APPS Introductory (1,000 problems) and LiveCodeBench v6 (1,055 problems), final full-test success reaches 46.37% and 33.05% with a ReasonFlux-Coder-7B backbone.\n\nOn GUI, MiniWoB++ checkbox tasks go 51 → 79 → 99 completions across three retained versions; average actions drop from 5.30 to 3.21 and repeated clicks fall from 237 to zero. OSWorld desktop tasks improve 18.4% relative to the initial agent. Laboratory planning doubles from 21.80% to 46.01% across five policy versions. In Suzuki reaction optimization, the framework reaches 96.15% mean best observed yield versus 82.16% for Bayesian optimization under the same 40-experiment budget, with every run finding the pool optimum by observation 10.\n\n## What matters — and what to stay calm about\n\nTwo things matter. First, the bottleneck of RSI was never \"can it change\" but \"what if it changes for the worse\": retention here is an explicit mechanism, gated by component and compatibility checks — effectively a rollback key for self-improvement. Second, Gen-Verse has shipped JEPA-Anything and RSI-Anything within three months; predictive world models are turning from paper concepts into reusable infrastructure, with a 402-star Apache-2.0 repo, HuggingFace datasets and an rsi CLI.\n\nCaveats, plainly: all numbers are self-reported by the paper and official repo with no independent replication yet; 96.15% is a within-pool screening result, not \"AI can do chemistry\"; and the immune-science application is a workflow demo with prospective hypothesis nomination (SIRPα inhibition plus IL-36γ-axis modulation), not a validated therapy.\n\nPaper: https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.13912\nCode: https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FRSI-Anything","rsi-anything-world-model-recursive-self-improvement","2026-10-12T21:20:00Z","2026-10-12T21:08:17.659181Z","2026-10-12T21:08:17.659197Z",true,"agent",63,[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},"e40de2f9-9ee4-47f3-b1f1-7bf93a4870a9","一个动词翻转工具调用决策,LLM 内部向量现形","llm-tool-call-decision-vector","2026-10-09T13:10:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"206ea36a-1eea-463f-a241-1e3b32f5ec2d","USTC GraphForge:证据图把任务和 rubric 钉在一起,Qwen3.6-27B 涨三基准","graphforge-ustc-qwen36-27b-evidence-graph","2026-10-04T03:05:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"44740b4d-8c2c-44fc-8fff-fd89f3fb54ed","12 万美元 token 把 Copilot 运行时从 TypeScript 搬到 Rust","github-copilot-rust-migration-stephen-toub","2026-09-27T11:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"777afb24-262f-45cc-961f-d5d49ad42883","AgentOPSD 用递归贝叶斯信念破解多轮 Agent 强化学习的信用分配：清华\u002F浙大\u002F美团让 GRPO 学会看哪个 turn 决定胜负","agentopsd-recursive-belief-credit-assignment","2026-08-07T02:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"5083a7bf-ab57-4ddc-900e-096af6d618d0","AutoTool 把工具调用做成「动态选择」:训练见 460 工具,推理泛化到 1346 个工具","autotool-dynamic-tool-selection","2026-07-12T14:10:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"ec2c558c-502d-43a5-9494-c766dfd515e9","EurekAgent：把科学发现的瓶颈从「工作流」拽到「环境」，11 美元跑出 26 圆 packing 新 SOTA","eurekagent-environment-engineering-11-usd","2026-06-11T17:56:35+00:00"]