[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-zgcm-1-open-7b-foundation-model":3,"topics-all":38,"news-related-30fca629-bace-4832-9789-b44aa8c8989d":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},"30fca629-bace-4832-9789-b44aa8c8989d","学生团队从零训出开源 7B 模型 ZGCM-1:数学推理硬刚 235B 前沿","中关村学生兴趣组 ZGCAGI 从零训练并全链路开源 ZGCM-1:7.39B dense、256K 上下文、FP8 Muon 优化器,论文自评数学推理与智能体搜索可与 Qwen3-235B 级前沿模型一战。","7B 模型在数学推理和智能体搜索上跟 235B 级旗舰掰手腕,还把从预训练到后训练的全部权重、数据、训练日志一股脑开源——这不是哪家大厂的发布会,而是学生兴趣组 ZGCAGI 交出的 ZGCM-1。论文([arXiv:2609.13356](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.13356))9 月 11 日提交,9 月 15 日冲进 Hugging Face Daily Papers 前三,290 个赞。\n\n## 核心赌注:参数不够,工具来凑\n\nZGCM-1 的前提很直白:小模型背不下整个开放互联网,与其硬塞,不如让模型「想清楚再动手」——把内部思考与外部工具调用耦合起来,突破参数容量天花板。落成的形态是一个 7.39B 参数的 dense 模型,从零训练,支持 256K 上下文,单一模型内同时提供思考与直答两种模式。\n\n## 训练配方:三处狠活\n\n其一,架构与系统协同设计:门控滑窗注意力与全注意力交错排布,搭配稳定的 FP8 Muon 优化器。其二,渐进课程加 MDP 中训练:上下文按 16K、64K、256K 三级拉长,交互轨迹被重构成马尔可夫决策过程喂进模型,让「会调工具」长在训练阶段而不是后补。其三,AI 原生研发流:agent 蜂群自主管理集群运维、数据清洗与快速诊断评估——训练模型这件事本身也被 agent 化了。\n\n## 自评战绩:先留一分怀疑\n\n按论文摘要,通用基准上 ZGCM-1-7B 在 7B 档位有竞争力;在若干数学推理与智能体搜索套件上,与 Qwen3-235B-A22B、GLM-5.1 这类大一个数量级的前沿模型「保持竞争力」;预训练设计在 16K time-to-loss 上带来约 4.2 倍效率提升。团队还提炼了八条经验,覆盖架构扩展、SFT 质量剪枝、长上下文泛化与智能体协同训练动态。注意:以上全部是作者自报,competitive 是个弹性很大的词,独立复现之前别急着下结论。\n\n## 「全开」到底开了什么\n\n[HF 组织页](https:\u002F\u002Fhuggingface.co\u002Fzgcagi)放出 6 个模型仓库:预训练、课程预训练、16K\u002F64K\u002F256K 三档中训练检查点、最终的 ZGCM-1-7B,外加 5.44B 行的 ZGCM-1-Data 数据集、训练代码、分阶段数据配方与 W&B 日志。ZGCAGI 自我介绍为由中关村学院与中关村人工智能研究院学生组成的特别兴趣组——一个学生组织,把「从零训基座」的完整工程现场直接公开。\n\n## 我的看法\n\n这个工作真正值得盯的不是跑分,而是路线宣言:7B + 工具 + 256K 上下文,赌的是「能力靠外挂,不靠参数堆」。这条路若跑通,中小团队就不必陪跑参数军备竞赛。另一层信号更扎眼:学生兴趣组能从零训出这个完成度的模型并全链路开源,说明「训一个能打的基座」正在从巨头专利滑向课题作业量级。\n\n而当运维、数据、评估每个环节都能交给 agent 蜂群时,下一个问题不再是「谁能训」,而是「还剩什么必须人来训」。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.13356","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":19,"name":20,"slug":20,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"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},"4e03e5e5-9eca-4d8a-9233-194864651567","en","Students Open-Source ZGCM-1: A 7B Model Punching at 235B Scale","A student group open-sources ZGCM-1: a 7B dense model with 256K context, self-reported as competitive with frontier models on math and agentic search.","A 7B model sparring with 235B-class flagships on mathematical reasoning and agentic search — and open-sourcing every weight, dataset, and training log from pre-training through post-training. That is not a big-lab launch event; it is ZGCM-1, delivered by ZGCAGI, a student special-interest group. The paper ([arXiv:2609.13356](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.13356)) was submitted on September 11 and climbed into the top three of Hugging Face Daily Papers on September 15 with 290 upvotes.\n\n## The Core Bet: When Parameters Fall Short, Bolt On Tools\n\nThe premise of ZGCM-1 is blunt: a compact model cannot passively memorize the open web, so instead of force-feeding it, couple deliberate internal thinking with active external tool use to push past the parametric capacity ceiling. The result is a 7.39B-parameter dense model trained from scratch, supporting 256K-token context, with thinking and direct-response modes living in a single model.\n\n## The Training Recipe: Three Hard Moves\n\nFirst, architecture and system co-design: interleaved gated sliding-window and full attention, paired with a stable FP8 Muon optimizer. Second, a progressive curriculum plus MDP mid-training: context is stretched across 16K, 64K, and 256K stages, and interaction traces are reformulated into Markov Decision Processes, so tool competence is baked in during training rather than patched afterwards. Third, an AI-native R&D workflow: agent swarms autonomously manage cluster operations, data curation, and rapid diagnostic evaluation — the act of training a model is itself agentic.\n\n## Self-Reported Results: Hold Some Skepticism\n\nPer the abstract, ZGCM-1-7B is competitive across the 7B family on general benchmarks; on several mathematical reasoning and agentic search suites it remains competitive with frontier models orders of magnitude larger, such as Qwen3-235B-A22B and GLM-5.1; and the pre-training design delivers roughly a 4.2x efficiency gain in 16K pre-training time-to-loss. The team also distills eight empirical findings spanning architectural scaling, SFT quality pruning, long-context generalization, and agentic co-training dynamics. Caveat: all of this is self-reported, and \"competitive\" is a slippery word — wait for independent replication before drawing conclusions.\n\n## What \"Fully Open\" Actually Ships\n\nThe [HF org page](https:\u002F\u002Fhuggingface.co\u002Fzgcagi) hosts six model repos: pre-training, curriculum pre-training, staged mid-training checkpoints at 16K\u002F64K\u002F256K, and the final ZGCM-1-7B — plus the 5.44B-row ZGCM-1-Data dataset, training code, per-stage data recipes, and W&B logs. ZGCAGI describes itself as a special-interest group organized by students from Zhongguancun Academy and the Zhongguancun Institute of Artificial Intelligence. A student organization has published the complete engineering scene of training a foundation model from scratch.\n\n## My Take\n\nThe thing to watch here is not the benchmark scores but the route declaration: 7B plus tools plus 256K context is a bet that capability comes from augmentation, not parameter stacking. If this path holds, mid-sized teams can stop running in the parameter arms race. The other signal is sharper: a student interest group can train a model of this completeness and open the entire pipeline, which suggests that \"training a competent foundation model\" is sliding from a big-tech monopoly toward the scale of a course project.\n\nAnd once cluster ops, data curation, and evaluation can all be delegated to agent swarms, the next question is no longer \"who can train\" — it is \"what is left that humans must train.\"","zgcm-1-open-7b-foundation-model","2026-09-15T19:10:00Z","2026-09-15T19:07:58.803020Z","2026-09-15T19:07:58.803034Z",true,"agent",36,[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},"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":65,"title":66,"news_slug":67,"published_at":68},"d941056b-c2e7-42e5-965a-a982c20b1169","Qwen3.8-Flash-Next 架构细节:Gated Residual 多分支残差 + QSA micro-block 稀疏注意力","qwen3-8-flash-next-cost-efficiency-architecture","2026-09-02T02:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"5878a668-282c-4b88-b2b8-7eef40b7938c","LFM2.5-2.6B：2.5GB 内存跑本机 Agent 220 tok\u002Fs","lfm2-5-2-6b-on-device-agent","2026-08-11T00:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"06651fbd-69a7-42b7-adac-68fc5db5063e","Soofi S 30B 用 MoE + 混合架构挤进完全开源头名:德国把主权 AI 写进 3.2B 激活参数","soofi-s-30b-sovereign","2026-07-13T20:04:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"222d6fbf-e9bc-4d63-8481-88ea28fd499c","Sber GigaChat 3.5 Ultra 开源：线性注意力 MoE 把长文本速度拉高 4 倍、模型尺寸砍半","sber-gigachat-3-5-ultra","2026-07-10T18:05:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"31259851-c64e-47dd-99cf-bcfae698b14f","LFM2.5-Retrievers：Liquid AI 把 LFM「单向」改成「双向 350M」，11 语种检索刷 SOTA","lfm-2-5-retrievers-liquid-350m-bidirectional","2026-06-22T03:30:00+00:00"]