[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-softwaregym-cua-synthesized-environments":3,"topics-all":38,"news-related-bd3f7140-0599-4dc7-b623-84a88432a53c":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},"bd3f7140-0599-4dc7-b623-84a88432a53c","Agent训练缺环境?上海交大合成182个软件来练","上海交大开源SoftwareGym:自动合成GUI软件环境训练电脑操作智能体,先生成多样UI设计,编码智能体实现为可运行应用,再经OSWorld虚拟机真实交互验证。论文报告合成12类182个应用,通过率91.5%;配套120任务基准,训练后OSWorld Pass@3从41.39%升到56.97%。","电脑操作智能体(computer-use agent)这两年进步飞快,但训练环境一直是瓶颈:真机应用构建贵、覆盖窄,浏览器环境交互深度又不够。上海交大 10 月 8 日放上 arXiv 的 SoftwareGym 给出一条自动化路线——不手工搭环境,而是流水线合成软件应用,再用真实 GUI 交互验证。\n\n## 两条老路的死胡同\n\n论文开篇把现状讲得很直白:原生应用环境需要大量构建与集成工作,应用覆盖因此受限,还容易让模型过拟合到少数应用上;浏览器托管环境好部署,但交互深度和工作流复杂度有限——点得动网页,练不了桌面软件。环境稀缺,直接卡住了电脑操作智能体规模化监督信号的供给。\n\n## 流水线:设计、实现、验证、返修\n\nSoftwareGym 把\"造软件\"本身变成一条多智能体流水线,代码已在 GitHub 开源。给定软件类型(如 ChatAPP、PicEditor),流水线四步走:\n\n1. **设计**:FunctionDesigner 写需求规格;StyleDesigner 与 SubStyleDesigner 产出多套互不相同的 UI 风格蓝图(Avalonia `.axaml`);IconDesigner 生成配套 SVG 图标集,带硬校验门;\n2. **实现**:Coder 把蓝图落成可运行应用(.NET 8 + Avalonia 11),DevCoder 负责评审和就地修复,所有设计编码环节都跑在 Codex CLI 上;\n3. **验证**:EvalTaskGenerateAgent 生成 GUI 测试用例,GUIEval 启动 OSWorld Docker 容器(真实 Ubuntu 虚拟机),由 GUI 智能体实际操作应用,输出每轮 summary.json;\n4. **返修**:v1 到 v3 循环,每轮把上轮评测摘要与评审意见回灌给 Coder\u002FDevCoder 增量修复。\n\n一个工程细节:SoftwareGym 直接复用 OSWorld 的 Docker 虚拟机基础设施和 Ubuntu 桌面,验证用截图加 PyAutoGUI 动作,保证训练环境与下游评测条件一致——这也解释了合成环境学到的能力为何能迁移。\n\n## 数字:182 个应用与一次域外迁移\n\n论文报告的规模是:合成出覆盖 12 个类别的 182 个应用,平均 GUI 任务通过率 91.5%;配套发布 SoftwareGym-Bench,一个同样横跨 12 类的 120 任务评测套件。关键结果在迁移上——用这些环境里收集的轨迹训练模型后,在严格域外的 OSWorld 基准上 Pass@3 从 41.39% 提升到 56.97%,在自家 SoftwareGym-Bench 上拿到 40.0%。\n\n## 该泼的冷水\n\n复现门槛不低:README 明说 OSWorld 虚拟机基础镜像约 35 GB 且不随包附带;需要自装 Codex CLI(锁定 @openai\u002Fcodex 0.156.1)、.NET 8 SDK 和 conda 环境;模型路由只认环境变量,设计编码阶段还要求端点支持 OpenAI Responses API。仓库目前 5 个 commit、16 stars,是典型的论文先行、工程刚落地状态。\n\n这条路线的想象空间在闭环:合成环境、收集轨迹、训练更强的智能体,再用更强的智能体去验证更多环境。当\"环境\"从稀缺资产变成可再生物料,电脑操作智能体的数据瓶颈才真正开始松动。论文与代码:[arXiv:2610.12611](https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12611),GitHub 仓库 [MrKiing-Jin\u002FSoftwareGym](https:\u002F\u002Fgithub.com\u002FMrKiing-Jin\u002FSoftwareGym)。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12611","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},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",{"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},"8d24e795-3ab1-42a7-be51-d344f42d82e9","en","SoftwareGym: 182 synthesized apps train computer-use agents","SJTU open-sources SoftwareGym, synthesizing 182 GUI apps to train computer-use agents; CUAs trained on them lift OSWorld Pass@3 from 41.4% to 57.0%.","Computer-use agents (CUAs) are improving fast, but the environments for training them remain a bottleneck: native applications are expensive to build and integrate, which narrows coverage and invites overfitting, while browser-hosted environments offer limited interaction depth. SoftwareGym, posted to arXiv on Oct 8 by a Shanghai Jiao Tong University team, takes an automated route: instead of hand-building environments, it synthesizes software applications along a pipeline and verifies they actually work through real GUI interactions.\n\n## Why the old paths dead-end\n\nThe paper is blunt about the status quo. Native application environments require substantial construction and integration effort, limiting application coverage and raising the risk of application-specific overfitting. Browser-hosted environments are easier to deploy but cap interaction depth and workflow complexity — you can click through web pages, but you cannot practice real desktop software. Scarce training environments directly throttle the supply of scalable supervision for CUAs.\n\n## The pipeline: design, build, evaluate, repair\n\nSoftwareGym turns \"making software\" into a multi-agent pipeline, with code open-sourced on GitHub. Given a software type (e.g. ChatAPP, PicEditor), the pipeline runs in four steps:\n\n1. **Design** — FunctionDesigner writes the requirement spec; StyleDesigner and SubStyleDesigner produce multiple distinct UI style blueprints (Avalonia `.axml`-style `.axaml`); IconDesigner generates per-style SVG icon sets with hard validation gates.\n2. **Build** — Coder implements the app (.NET 8 + Avalonia 11); DevCoder reviews and fixes in place. All design and coding stages run through the Codex CLI.\n3. **Evaluate** — EvalTaskGenerateAgent creates GUI test cases; GUIEval boots OSWorld Docker containers (real Ubuntu VMs) and has a GUI agent actually operate the app, producing a per-round summary.json.\n4. **Repair** — rounds v1 to v3, each feeding the previous eval summary and review back to Coder\u002FDevCoder for incremental fixes.\n\nOne engineering detail worth noting: SoftwareGym reuses OSWorld's Docker-based VM infrastructure and Ubuntu desktop, validating via screenshots and PyAutoGUI actions, so training conditions stay consistent with downstream evaluation — which helps explain why capabilities learned in synthesized environments transfer.\n\n## The numbers: 182 apps and an out-of-domain jump\n\nThe paper reports 182 synthesized applications across twelve categories with a mean GUI task pass rate of 91.5%, plus SoftwareGym-Bench, a 120-task evaluation suite spanning the same categories. The key result is transfer: after training on trajectories collected from these environments, the model improves Pass@3 from 41.39% to 56.97% on the strictly out-of-domain OSWorld benchmark and scores 40.0% on SoftwareGym-Bench.\n\n## Caveats worth noting\n\nThe reproduction bar is nontrivial. The README states the OSWorld VM base image (~35 GB) is not included in the package; you must install the Codex CLI (pinned to @openai\u002Fcodex 0.156.1), the .NET 8 SDK, and a conda environment; model routing reads environment variables only, and the design\u002Fcoding stages require an endpoint that supports the OpenAI Responses API. Seed materials for media\u002Ffile-type software must be self-supplied, and container cleanup is manual. The repository currently sits at 5 commits and 16 stars — a typical paper-first, engineering-just-landed state.\n\nThe imaginative part of this route is the loop: synthesize environments, collect trajectories, train stronger agents, then use stronger agents to validate more environments. When environments shift from scarce assets to regenerable material, the data bottleneck for computer-use agents finally starts to loosen. Paper and code: [arXiv:2610.12611](https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12611), GitHub [MrKiing-Jin\u002FSoftwareGym](https:\u002F\u002Fgithub.com\u002FMrKiing-Jin\u002FSoftwareGym).","softwaregym-cua-synthesized-environments","2026-10-12T19:08:19Z","2026-10-12T19:08:21.795202Z","2026-10-12T19:08:21.795213Z",true,"agent",196,[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},"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":65,"title":66,"news_slug":67,"published_at":68},"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",{"id":70,"title":71,"news_slug":72,"published_at":73},"aa718f3c-14cf-4984-8762-794eaaee8c16","MAI-Gym:9B 智能体在 2 万个并行环境里做在线 RL","mai-gym-gui-agent-online-rl-9b","2026-10-21T09:30:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"bb6e0f7b-3899-4e45-a267-bfa512d470b0","FDE-Gym:95个真实交付任务,最强Agent只过14个","fde-gym-agent-delivery-benchmark","2026-10-18T23:08:06+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"48726cf0-6f0c-4b09-841e-06aa8b0ffcfd","上海AI Lab Intern-Decision:多模态决策模型,4B 微超 Jev 90.02%","intern-decision-qwen3-5-multimodal-decision-model","2026-10-18T11:13:06+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"da6cbb98-3b89-4631-8dbd-59411136fb30","哈佛用 700 家公司拆解 AI 编程智能体:30% 代码增量,49% 合并时延","harvard-coding-agent-review-bottleneck","2026-10-14T11:30:00+00:00"]