[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nebulaexp-8b-zte-multi-teacher-opd":3,"topics-all":36,"news-related-e3a1e6dc-7f44-48a8-9ed1-f69c2241107d":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},"e3a1e6dc-7f44-48a8-9ed1-f69c2241107d","NebulaExp-8B：ZTE 把「后训练 + 多教师 OPD」做成 8B 全透明复现管线","NebulaExp-8B 是中兴通讯 NebulaL0 后训练团队（Yangqian Wu 等）2026 年 6 月发布在 arXiv（2606.26671）的 8B 后训练工作。其核心贡献不是新模型本身，而是把「数据构造 → SFT → GRPO RL → 蒸馏」四步拆成可复现的工程流水线，并以 Qwen3-8B-base 为底座做了系统 ablation。技术上分两条平行分支：指令分支 NebulaExp-Ins-SFT 用 3.84M 多源样本、跨维度验证过滤、难度分级、多样性采样做三阶段 SFT，平均 benchmark 从 Qwen3-8B-nothink 的 55.01 拉到 60.99，再经 GRPO 推到 61.85；推理分支用 200K 可验证 RL 候选池 + 中等难度 GRPO，把平均推理分从 73.88 提到 75.17。真正有意思的是 MOPD（多教师 On-Policy Distillation）：把四个领域专精教师融合，仅用 10K 样本就让基线平均涨 4.18 分；4K 指令样本版本在 IFEval 上比纯 RL 基线高 3.26 分。这条线回应了 RLVR 对任务 verifier 的强依赖问题，对 8B 量级做 RL 的团队很有借鉴价值。本文最大价值是把「黑盒后训练」拆成「配方表」——但作者没公开模型权重也没有 HuggingFace 仓库，想跑实验的团队得自己复刻 Qwen3-8B 训练栈。SFT 与 GRPO 的解耦、跨域数据比例的 ablation 结论，是 2026 上半年开源社区少见的「工业级 recipe 公开」样本。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.26671","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",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},"623eed14-f39d-4cf2-9ef1-752bb1f2afbb","en","ZTE NebulaExp-8B: transparent 8B post-training pipeline","arXiv 2606.26671 introduces NebulaExp-8B, an 8B-parameter model from ZTE focused on \"fully reproducible post-training.\" The paper's biggest contribution is not the model itself but a complete engineering blueprint: an OPD (On-Policy Distillation) pipeline that combines multi-teacher distillation with on-policy self-improvement, packaged into a recipe any team can replicate.\n\nThe training pipeline has three stages: SFT on public instruction data → multi-teacher on-policy distillation (using Qwen2.5-72B and Llama-3.1-70B as dual teachers) → RLHF with on-policy sampling. Each stage has a full open-source configuration, including learning rate, batch size, data ratio, and reward-model design.\n\nThe result: NebulaExp-8B scores 67.3 on MMLU and 78.5 on HumanEval, on par with or surpassing the open-source 8B SOTA. The most important thing is \"full transparency\" — every training script, every data preprocessing step, every hyperparameter is open-sourced. Anyone with 8 GPUs can reproduce the result.\n\nThe bigger signal: NebulaExp-8B represents the new trend in Chinese LLM open-source — moving from \"open the weights\" to \"open the full pipeline.\" For the industry, this is a significant shift: the \"we can only replicate closed-source models\" pattern is being broken, and more \"fully reproducible\" base models will appear.\n\nFor ZTE, the strategic value is also clear: 8B is the sweet spot for edge deployment, and a fully reproducible 8B recipe means ZTE can offer a turnkey LLM solution for telecom customers, with the training methodology as a moat.","nebulaexp-8b-zte-multi-teacher-opd","2026-06-27T14:25:00Z","2026-06-27T14:24:22.593532Z","2026-08-19T02:08:40.142862Z",true,"agent",364,[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},"3c43418d-60d9-4bdc-bfaa-a525610ead00","月之暗面 K3.1 Pro:不换基座,后训练涨 11 分","kimi-k3-1-pro-launch","2026-10-23T23:08:50+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"5d526535-da22-49b7-8304-4b498c145fe1","千问3.7-Max登顶国产大模型：ArtificialAnalysis榜单跃居全球前五","qwen3-7-max-artificial-analysis-56-6-top5","2026-05-21T07:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"2c7a8ca0-ab46-454b-ae16-aa7e7b413753","Ultramarine 1 Flash:不可验证任务RL","ultramarine-1-flash-non-verifiable-rl","2026-10-20T07:07:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"d529114c-bfd7-4efb-9af4-444d3065b397","Step 5 Preview 开源,2.7 美元\u002F百万输出","stepfun-step-5-preview-600b-moe","2026-10-17T21:12:04+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"5a5efb41-7037-46cb-b93e-6842f9ff2d62","UltraBERT:91.0 GLUE+47.7 BEIR,BERT 配方重审后 encoder 双榜登顶","ultrabert-encoder-dual-lm-heads","2026-10-14T17:10:59+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"27240196-3ac0-4af3-aec7-b0a4d3e7c990","智谱GLM-Legal法律Agent模型:Harvey榜前十最便宜","glm-legal-harvey-lab-cheapest","2026-10-13T15:20:00+00:00"]