[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-traceforge-family-native-distillation":3,"topics-all":38,"news-related-3a419e1c-fb2b-467e-a5b6-a7188525bc03":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},"3a419e1c-fb2b-467e-a5b6-a7188525bc03","TraceForge:推理痕迹自己造,学生反超教师真迹","偷加密思维链违反服务商条款,TraceForge 训练生成器用学生家族自己的风格补写推理痕迹。论文报告 Qwen3.5-9B 用家族原生痕迹训练,HumanEval+ 与 MBPP+ 分别比用教师 Fable 5 真迹高 3.7 与 7.1 分,比直接蒸馏 Qwen3.8-Max 高 21 与 35 分。","闭源实验室把推理痕迹锁起来已有一段时间:API 只返回答案加一小段总结,思维链本身不再给。用加密块跨模型重放能把思维链偷回来——今年 8 月的 Stealing Reasoning Traces 论文证明过这条路可行——但它违反服务商条款,没有产品能长期押在上面。10 月 21 日,独立研究者 Yan Wang 在 arXiv 发布 TraceForge,给出第三条路线:不偷教师的推理,用学生模型家族自己的风格补写一份。\n\n## 推理痕迹可以自己造\n\nTraceForge 让两个模型对着长大。监督阶段,生成器从上下文和答案还原 Qwen3.8-Max 的约一万条真实推理痕迹;对抗阶段,判别器学习分辨真假,骗过判别器的样本成为下一轮的训练数据,五轮在线共同训练后收敛。推理时,生成器给教师的答案配一份「家族原生」(family-native)推理痕迹,判别器再从四个候选里挑最像真的那个。论文用闭源裁判 GPT-6.1-SOL 打分:合成痕迹的真实度拿到 68%,略高于真实痕迹的 66%,四选一后升到 82%。方法节里还记录了一个教训:作者原以为小模型能批大模型的作业,几百步后判别器就学歪了,最终判别器必须不小于生成器。\n\n## 五个语料的干净对照\n\n学生统一为 Qwen3.5-9B-Base,底座、训练配方、预算全部固定,数据统一来自 5,000 条带 Fable 5 真实推理痕迹与答案的 Claude Code 轨迹(3,500 个问题、256K 上下文窗口、八张 H20 GPU),只换推理痕迹和答案的来源。论文报告的最终成绩:用家族原生痕迹替换 Fable 5 自己的痕迹,HumanEval+ 从 64.63 涨到 68.29,MBPP+ 从 52.38 涨到 59.52,分别提升 3.7 与 7.1 分;直接蒸馏家族里最强的 Qwen3.8-Max,反而只有 46.95 和 24.60,落后 21 和 35 分;对照组 TraceInversion 的通用合成痕迹崩到 3.66 和 6.61。换第二位教师 GPT-6.1-SOL 后结论一致:同样的答案,自家痕迹拿到 59.15 和 60.05,第三方合成还是 3.66 和 6.61。\n\n## 为什么自家口音更好\n\n论文把原因归于灾难性遗忘更小:学生与痕迹同族,吸收时不必先忘掉自己的习惯。旁证有两个:用教师真迹训练的 S1 全程不稳定,HumanEval+ 在 66.46、33.54、64.63 之间震荡,一次 33 分的坠落又自行修复;而 Qwen-Max 组与 TraceInversion 组的输出越训越多死循环与重复,也就是圈内说的「吸二手烟」效应,家族原生痕迹完全没有出现这个模式。教师能力是必要条件,但不是充分条件——答案可以来自最强模型,表达方式得是学生自己的。\n\n## 边界与评论\n\n需要诚实标注的:单作者独立研究,每个格子只跑一次,裁判是闭源模型,5,000 条轨迹的来源保密。但五个语料、两位教师、逐 checkpoint 的交叉对照,在这个方向上是少见的干净实验。对行业,它把推理痕迹的获取战从「偷」挪到了「造」:合法的代价第一次不是性能。论文还给出零 Claude\u002FOpenAI token 参与反向传播的变体——痕迹配 Qwen3.8-Max 写的答案——成绩与主实验基本持平,法律暴露面进一步清零。回到最初的问题:最好的学生不是最强的老师教出来的,而是最强的老师用学生最熟悉的方式教出来的。原文见 arXiv:2610.29680(https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.29680)。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.29680","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},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",{"id":22,"name":23,"slug":23,"description":14,"color":14},"c187600e-804c-4697-b828-1e4330e0eb10","qwen",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"8bf976d3-c27d-482f-b407-a7baca1644a7","en","TraceForge: Self-Written Traces Beat the Teacher's Own","TraceForge writes family-native traces for a teacher's answers; Qwen3.5-9B reportedly beats training on Fable 5's real traces by 3.7 and 7.1 points.","Closed labs have been locking reasoning traces away for a while: the API returns an answer plus a short summary, and the chain of thought itself stays hidden. Replaying encrypted blocks across models can steal it back — the Stealing Reasoning Traces paper from this August proved the route works — but it breaks provider terms of service, so no product can build on it long term. On October 21, independent researcher Yan Wang published TraceForge on arXiv, offering a third route: do not steal the teacher's reasoning; write your own, in the style of the student's own model family.\n\n## Traces can be synthesized\n\nTraceForge grows two models against each other. In the supervised stage, a generator learns to restore roughly ten thousand real reasoning traces of Qwen3.8-Max from their context and answer. In the adversarial stage, a critic learns to tell real traces from generated ones; samples that fool the critic become the next round's training data, and online co-training converges after five rounds. At inference, the generator writes a \"family-native\" trace for the teacher's answer, and the critic picks the most convincing of four candidates. Scored by a closed judge (GPT-6.1-SOL), synthetic traces reach 68% perceived authenticity, slightly above the 66% of real traces, and 82% with best-of-four selection. The method section records one honest mistake: the author assumed a smaller model could grade a larger one; within a few hundred steps the critic fitted shortcuts, so the critic must be at least as large as the generator.\n\n## A clean five-corpus comparison\n\nEvery student is fine-tuned from Qwen3.5-9B-Base with base model, recipe, and budget held fixed. The data pool is 5,000 Claude Code trajectories carrying Fable 5's real reasoning traces and answers (3,500 questions, 256K context window, eight H20 GPUs), and only the source of the trace and the answer varies. The paper reports: replacing Fable 5's own traces with family-native ones lifts HumanEval+ from 64.63 to 68.29 and MBPP+ from 52.38 to 59.52 — gains of 3.7 and 7.1 points; distilling directly from the family's strongest model, Qwen3.8-Max, yields only 46.95 and 24.60 — 21 and 35 points behind; the generic TraceInversion traces collapse to 3.66 and 6.61. Switching to a second teacher, GPT-6.1-SOL, keeps the conclusion: given identical answers, family-native traces reach 59.15 and 60.05 while the third-party synthesizer stays at 3.66 and 6.61.\n\n## Why the family accent wins\n\nThe paper attributes the gain to smaller catastrophic forgetting: student and traces come from the same family, so the student absorbs them without unlearning its own habits. Two pieces of side evidence: the teacher-trace corpus (S1) is unstable, with HumanEval+ swinging between 66.46, 33.54, and 64.63 — a 33-point drop that repairs itself; and the Qwen-Max and TraceInversion corpora drift into dead loops and repetition as training proceeds, the so-called second-hand-smoke effect, a pattern family-native traces never show. Teacher capability is necessary but not sufficient: the answer can come from the strongest model, but the delivery has to be the student's own.\n\n## Limits and take\n\nHonest caveats: single-author independent research, one run per cell, a closed-source judge, and a confidential source for the 5,000 trajectories. Still, five corpora, two teachers, and per-checkpoint cross-checks make this an unusually clean experiment for the direction. For the industry, it moves the fight over reasoning traces from stealing to synthesizing: for the first time, staying legal costs no performance. The paper also offers a variant where zero Claude or OpenAI tokens are ever backpropagated — the family-native trace paired with a Qwen3.8-Max answer — which performs essentially the same as the main setup, removing the legal exposure entirely. Back to the opening question: the best student is not produced by the best teacher alone, but by the best teacher teaching the way the student learns best. Paper: arXiv:2610.29680 (https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.29680).","traceforge-family-native-distillation","2026-10-22T15:10:42Z","2026-10-22T15:10:51.060639Z","2026-10-22T15:10:51.060653Z",true,"agent",12,[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},"d1b7ef56-357c-4b38-841e-4d4d60970490","Waypoint-OPD给蒸馏加方向门控：教师前后打架就不学","waypoint-opd-direction-gated-distillation","2026-10-17T23:20:00+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},"fd29c2f1-175f-416e-bde2-95dd65ca3465","KronLSTM 精准改写记忆:线性注意力长上下文新解","kronlstm-kronecker-delta-rule","2026-10-20T15:06:16+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"559df680-f13e-48f0-9263-73116aaff411","上交大DLite免历史投机解码:Agent提速2.74倍","dlite-history-free-speculative-decoding","2026-10-19T17:07:52+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},"eeb3f139-8df8-40d7-a1e6-82e49b668547","VideoJev:视频问答提速100倍,不生成只打分","videojev-system-one-video-understanding","2026-10-14T15:09:17+00:00"]