[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-agentgarten-code-worlds-for-evolving-agents":3,"topics-all":38,"news-related-d5ba4b97-e08e-4ddb-80d8-70f8060f6d9a":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},"d5ba4b97-e08e-4ddb-80d8-70f8060f6d9a","AgentGarten:把世界写成代码,智能体4轮学会躲猫猫","MirroS-Lab 在 arXiv 发布 AgentGarten:模拟器与游戏引擎维护状态和规则,共享神经渲染器把深度图和法线变成智能体看到的画面。智能体 4 轮经验即学会核心策略,对照组 RL 需要百万轮。渲染器代码和权重已开源,训练循环未放出。","智能体研究里有个不太起眼却很硬的约束:模型能学到什么,取决于它在什么环境里练。环境的状态、规则、动力学要自洽,画面还得像真实世界——这两条同时满足,恰恰是行业瓶颈。MirroS-Lab 10 月 8 日放在 arXiv 上的 AgentGarten 给了一个相当漂亮的解法,在 Hugging Face 10 月 9 日的论文日榜排到第一,目前 141 个赞。\n\n## 架构:代码管规则,渲染器管画面\n\nAgentGarten 把\"世界\"拆成两层。底层是模拟器和游戏引擎:维护持久的世界状态,执行程序定义的交互规则,保证物理上自洽。上层是一个共享神经渲染器:每个世界通过统一接口导出深度图和表面法线,渲染器据此生成智能体\"看到\"的画面。规则交给代码,观感交给神经网络,两层各司其职。\n\n渲染器不是从零训练的。团队拿 NVIDIA 的 Cosmos3-Nano 视频模型做底座,分三步改造:先做双向 rectified-flow 训练,再上自回归 block-causal 训练(带 KV cache 采样),最后用他们提出的 Adversarial Forcing 蒸馏成少步学生模型——每段回放精确可微,后面块的损失能反过来更新模型对历史的编码方式,再加真实数据判别器压制视觉质量短板。推理端逐块流式生成,手写 Triton kernel 加 CUDA graph,明确不用 torch.compile。\n\n## 效果:4 轮 vs 百万轮\n\n论文给出的对比相当悬殊:智能体只需要 4 轮经验就能学会核心策略,而传统强化学习对照组是百万轮量级。按论文作者在 Hugging Face 页面的描述,在捉迷藏任务里,躲藏方第 4 轮开始搭掩体,寻找方第 10 轮学会用坡道。机制上,每轮经验会被蒸馏成 playbook,后续智能体继承并继续改进——经验以可读文本的形式跨代累积,而不是全部堆进权重里。\n\n更关键的是扩展性:新世界直接用代码写,过同一个渲染接口就能跑,环境的数量和难度可以跟着智能体一起涨。\n\n## 开源了什么,还欠什么\n\n现在放出来的是 Apache-2.0 协议的渲染器训练与推理代码,权重挂在 Hugging Face,仓库发布三天攒了 143 个 star。但 README 的 TODO 里有两项还没勾:实时渲染引擎(流式服务器+交互前端),以及最核心的\"代码世界+智能体训练循环\"本身。也就是说,这套体系目前只交付了渲染器这一半,论文里那个 4 轮出师的完整闭环,外界还跑不起来。\n\n所以,把它当成\"世界模型当训练场\"路线的一个高质量组件来看,比当成完整系统更准确。代码世界能不能像论文设想的那样随智能体一起扩展,要等训练循环落地才知道。论文原文见 [arXiv:2610.12374](https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12374),代码在 [GitHub](https:\u002F\u002Fgithub.com\u002FMirroS-Lab\u002FAgentGarten)。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12374","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},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",{"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},"6b8be49e-e864-4800-ad4c-2eba9644f633","en","AgentGarten: Worlds as Code, Agents Learn in 4 Rounds","MirroS-Lab's AgentGarten couples simulators with a shared neural renderer; agents learn in 4 rounds vs millions for RL. Renderer code is open.","There's an unglamorous but fundamental constraint in agent research: what a model can learn is bounded by the environment it practices in. The environment needs faithful state, rules and dynamics, and realistic observations at the same time — and hitting both across diverse worlds has been the bottleneck. AgentGarten, released on arXiv on October 8 by MirroS-Lab, offers a clean answer. It hit #1 on Hugging Face's daily paper ranking for October 9 with 141 upvotes.\n\n## Architecture: code owns the rules, a renderer owns the pixels\n\nAgentGarten splits a \"world\" into two layers. The bottom layer is simulators and game engines: they maintain persistent world state and execute program-defined interaction rules. The top layer is a shared neural renderer: every world exports depth maps and surface normals through a common interface, and the renderer generates what the agent actually sees. Rules go to code, visuals go to a neural network.\n\nThe renderer is not trained from scratch. The team adapts NVIDIA's Cosmos3-Nano video model in three stages: bidirectional rectified-flow training, autoregressive block-causal training sampled with a KV cache, and finally their proposed Adversarial Forcing, which distills the model into a few-step student. Rollout replay is exact and differentiable, so losses on later blocks update how the student encodes history, while a real-data discriminator polishes visual quality. Inference streams block by block with hand-written Triton kernels and CUDA graphs — no torch.compile.\n\n## Results: 4 rounds vs millions\n\nThe paper's headline number is stark: agents learn core strategies from just 4 rounds of experience, versus millions of rounds for a conventional reinforcement-learning counterpart. In hide-and-seek, per the authors' note on the Hugging Face page, hiders build shelters by round 4 and seekers use ramps by round 10. After each round, experience is distilled into playbooks that later agents inherit and refine — knowledge accumulates as readable text across generations rather than piling into weights.\n\nEqually important is scalability: new worlds are written as code and rendered through the same interface, so environments can scale in number and difficulty alongside their agents.\n\n## What's open-sourced, and what's still owed\n\nWhat's available now is the renderer's training and inference code under Apache-2.0, with weights on Hugging Face; the repo has collected 143 stars in three days. But two items in the README's TODO remain unchecked: the real-time rendering engine (streaming server and interactive frontend) and — crucially — the code worlds and the agent practice loop itself. In other words, only the renderer half of the system has shipped; the full closed loop behind that 4-round result can't yet be reproduced outside the team.\n\nSo treat it as a high-quality component of the \"world models as training grounds\" route rather than a complete system. Whether code worlds can really scale alongside agents, as the paper envisions, will only be answerable once the practice loop lands. The paper is at [arXiv:2610.12374](https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12374), code on [GitHub](https:\u002F\u002Fgithub.com\u002FMirroS-Lab\u002FAgentGarten).","agentgarten-code-worlds-for-evolving-agents","2026-10-10T17:10:00Z","2026-10-10T17:10:52.287774Z","2026-10-10T17:10:52.287788Z",true,"agent",46,[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},"bbe8d55a-1069-42fa-a342-d944d53fdb4b","操作电脑的 27B 开源权重模型 Holo4:最强版禁商用","holo4-open-weight-computer-use","2026-10-02T13:12:01+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"dc61debd-1d77-4d5a-9d59-5b23c3da07de","蚂蚁开源Realtime-Venus：9B全双工模型边说边干活，三项续聊指标超GPT-4o","ant-realtime-venus-full-duplex-delegation","2026-09-30T23:10:53+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"af21d26d-d9cd-4526-ac4a-66366a45848c","AV-GRPO:8张A800给22B音视频模型做RL后训练","av-grpo-audio-video-diffusion-rl","2026-09-27T17:08:13+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"9210c86b-3e1d-434b-bbfd-78e62c698aed","WorldCrafter 开源:给视频世界模型装上可查询的 3D 记忆,转一圈回来还是那个房间","worldcrafter-video-world-model-3d-memory","2026-09-22T19:08:48+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"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":85,"title":86,"news_slug":87,"published_at":88},"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%","embodiedskills-vla-verify-loop","2026-09-08T17:10:00+00:00"]