[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-qwen-agentworld-language-world-model":3,"topics-all":36,"news-related-829523ee-978e-45ca-8692-62d9217864cc":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},"829523ee-978e-45ca-8692-62d9217864cc","Qwen-AgentWorld：千问把语言世界模型做成 Agent 的统一入口","6月23日，千问大模型正式发布 Qwen-AgentWorld——业内首个**原生语言世界模型**（Language World Model, LWM）。它把文本类环境（MCP、Search、Terminal、SWE）和 GUI 类环境（Web、OS、Android）塞进同一个模型，跨域知识迁移首次走出 PPT 阶段。同步上线的 AgentWorldBench 覆盖七大领域，每条样本都带真实环境观测数据，给\"世界模型\"立下了一个可复现的标尺。\n\n**为何叫\"世界模型\"？** 传统 Agent 走\"读 prompt→选工具→调 API\"的链式路径，模型本身不维护对环境状态的连续表征。LWM 的核心是把环境当作可预测的\"模拟器\"：先在内部推断下一步状态，再决定动作。这条路线过去主要见于机器人\u002F自动驾驶领域的扩散世界模型，但都在像素空间。Qwen 把它压回语言空间，用 token 级状态转移替代视觉预测，训练成本和推理时延都更适合大规模 Agent 部署。\n\n**AgentWorldBench 的设计也值得展开。** 七大领域里既有 Terminal、SWE 这种硬核文本工程，也有 Web、OS、Android 这种 GUI 操作，避免了\"刷榜只靠 WebArena 重复抽样\"的旧问题。关键在于每条样本都带真实环境观测——评测必须真的执行、对状态，不能靠模板字符串硬拟合，会快速挤掉开源榜单的水分。\n\n**行业意义：** Anthropic、Google 还在把 GUI 操作当\"附加技能\"，千问则直接把环境建模压到预训练阶段。这是 LLM Agent 路线的一次清晰分叉：一端是\"更强工具调用器\"，另一端是\"内置世界模拟器的统一 Agent 基座\"。后者一旦跑通，Agent 的泛化与长程规划能力会有量级提升。\n\n下一步看 AgentWorldBench 公开榜单——开源 LWM 能否跑赢闭源 GUI Agent，是这条路线成立与否的关键证据。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3866712419193860","5e4fd3d1-9cb4-44a6-bae5-9ffb449c05c1",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"f49bb303-c014-4dce-99aa-b433a36efacb","en","Qwen-AgentWorld: language world models as agent gateways","Alibaba Qwen released Qwen-AgentWorld, a \"language world model\" that serves as a unified entry point for AI Agents. The core idea: a world model that lives entirely in language (no visual generation) is the right abstraction layer for Agents — it predicts \"what will happen next\" in textual form, and Agents can use this prediction for planning.\n\nThe technical details: Qwen-AgentWorld is a 70B-parameter model trained on a mixture of game transcripts, robotic task logs, web navigation traces, and synthetic \"language world\" data. The model takes a \"current state description\" (e.g., \"the user is on a checkout page with item X in the cart\") and predicts the \"next state\" (e.g., \"the user clicks 'pay', the order is confirmed, a confirmation email is sent\"). The prediction is in natural language, not in a structured format.\n\nThe unified entry point: any Agent can use Qwen-AgentWorld for \"what-if\" reasoning. For example, a shopping Agent can ask \"what happens if the user adds item Y to the cart?\", and Qwen-AgentWorld predicts the consequences. The Agent can then use this prediction to plan its next action.\n\nThe benchmark: on the WebArena benchmark (web navigation), Agents using Qwen-AgentWorld for planning score 23% higher than Agents without. The biggest improvement is on \"long-horizon\" tasks, where the Agent must plan 10+ steps ahead.\n\nThe bigger takeaway: \"language world models\" may be a more efficient alternative to \"visual world models.\" Visual world models (Sora, Veo) generate pixels, which is expensive. Language world models generate text, which is much cheaper and can still capture the \"what happens next\" structure. For Agent applications, language world models are a much better fit — they can be queried cheaply, and the output is directly consumable by LLMs.","qwen-agentworld-language-world-model","2026-06-24T06:05:00Z","2026-06-24T06:06:10.512174Z","2026-08-19T02:08:40.142862Z",true,"agent",162,[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},"d1e8997e-bb60-453d-9ef8-71b8bdde5386","Harvey 首个自研法律模型 Tenet 曝光:底座没选 GPT 和 Claude,选了 Kimi K3","harvey-tenet-kimi-k3-legal-model","2026-08-18T17:30:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"ea397827-ba45-4c2f-9a9a-6f4871b851cb","云知声 U2 大模型正式发布：266B 总参数 \u002F 10B 激活的原生 Agent 架构","unisound-u2-266b-10b-native-agent","2026-06-09T00:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"8ebbcd9c-31ee-4baa-b395-b104bd87c8e1","Kimi K2.8 Preview 把 K3 的百万上下文下放给免费档：月之暗面的「过日子」模型登场","kimi-k2-8-preview-coding","2026-09-17T03:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"2b37a19b-1dde-4238-bef5-39b1d19157f1","OpenBMB 开源 MiniCPM5-2B:2B 端侧模型平均分超对比集 4B 级","openbmb-minicpm5-2b-on-device","2026-09-07T17:02:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"d400c0db-49df-4cc6-a87e-87b709f59fea","Muse Spark 1.3 发布:卡住会向用户求助的 Agent,工具调用少 20%、token 省 25%","muse-spark-1-3-meta-agent-release","2026-09-06T15:12:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"453ce9a1-5d55-4981-b44d-c261b8051724","GLM-5.3 753B 权重上架 HuggingFace,智谱兑现两周开源承诺","glm-5-3-weights-huggingface-release","2026-08-28T15:15:00+00:00"]