[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-google-opst-data-free-self-training":3,"topics-all":38,"news-related-edc863dc-1918-4fdf-8306-640d913e15e2":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},"edc863dc-1918-4fdf-8306-640d913e15e2","Google OPST:零人类习题,模型自出题自提升","Google 团队在 arXiv 提出 OPST 自训练框架:只给粗粒度主题描述,模型自己合成问题、特权推理示范与课程,配伪验证过滤器筛题,全程零人类标注问答。文本基准与 VQA 均取得一致自我提升,未训练域能力保留,代码暂未放出。","后训练阶段的每一分提升,几乎都拴在人类生产的监督信号上:要么是标准问答对,要么是人工策展的题集,要么是更强的教师模型。10 月 23 日,一支 Google 团队把论文《On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models》挂上 arXiv(编号 2610.32770),给出了另一个极端的答案:把这些外部依赖全部拿掉,只给模型几个粗粒度的主题词,让它自己给自己出题、自己批改、自己升级。\n\n## 把出题权交给模型\n\n论文的起点是一个已经颇为成熟的方向:on-policy 自蒸馏——让模型在不依赖更强外部教师的情况下做后训练。但作者指出,现有做法仍然需要 ground-truth 问答对、人工策展题集或其他监督信号,「人类数据」这根拐杖没有真正扔掉。\n\n这篇论文的框架(下称 OPST)把拐杖抽得更彻底。整个训练管线只需要一样人类输入:粗粒度、高层的主题描述。围绕这些主题词,模型自己合成问题与配套的「特权推理示范」(privileged reasoning demonstrations),自己搭出一条课程(curriculum)——从出题、解题到筛选,全程没有一道人类写的习题或答案。\n\n## 两道防自欺的闸门\n\n自己给自己出题,最大的风险是把错题当真题。OPST 的解法是一个伪验证过滤器(pseudo-verification filter):一道自生成的题目想留在课程里,必须满足一个条件——模型换一个全新上下文、换一种问题改写(paraphrase)之后,仍然能复现出原来的答案。答得出来,才算数。\n\n另一道闸门是「特权推理示范」本身。作者在 Hugging Face 论文页的讨论区里把评估范围说得更直接:他们看的不只是训练域内的自我提升,还包括混合多个生成域时,模型在未参与自训练的域上知识是否保留。结论是两头都成立:域内多个方向提升,域外能力不丢——这对长期困扰微调的「灾难性遗忘」是个正面信号。\n\n## 文本之外,还能看图\n\n框架的通用性不止于纯文本。论文展示了一个多模态版本 Visual OPST,把同一套自训练逻辑迁移到视觉输入上,在视觉问答(VQA)任务上取得一致的自我提升。作者在讨论区的总结是:OPST 在不使用任何人类策展问答的前提下,同时改善语言与多模态推理,并优于其他替代训练策略。\n\n团队阵容六人:Leon Sick、Lukas Hoyer、Yannick Strümpler、Christian Reisswig、Timo Ropinski、Federico Tombari,论文页面挂 Google 机构,10 月 26 日由一作本人提交到 Hugging Face Daily Papers,目前已有 11 个 upvote。\n\n## 论文先行,代码未到\n\n还有一个事实值得注意:截至发稿,Hugging Face 论文页显示「引用本文的模型 0、数据集 0、Space 0」——方法已经公开,工件还没落地。论文的结论也写得相当克制:它证明的是基础模型本身就具备丰富的潜在能力,不需要外部任务监督或策展题集分布,也能被放大出来。\n\n这件事的行业意味可以往两个方向读。乐观读法:合成数据的自我供给闭环又往前走了一步,数据生产从「人力瓶颈」变成「算力瓶颈」,对攒不起题库的小团队尤其友好。谨慎读法:伪验证过滤器筛掉的是「不一致」,而不是「不正确」——一道题如果模型在每次全新上下文里都稳定地答错,这个过滤器拦不住它。自训练放大的是模型已有的分布,顺带也会放大已有偏差,这个敞口论文没有完全合上。\n\n一个值得留在评论区的问题:当出题人、答题人和批卷人是同一个模型时,你愿意把多少后训练交给它?","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.32770","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"id":19,"name":20,"slug":20,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":22,"name":23,"slug":23,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"bcdec3f3-d3a7-48f7-8629-8ebf58690559","en","Google OPST: LLMs Write Their Own Training Curriculum","Google's OPST: models self-generate questions and curriculum from topic descriptions alone; consistent gains on language and VQA, code not yet out.","Post-training gains have always been tied to human-produced supervision: ground-truth question-answer pairs, human-curated problem sets, or a stronger teacher model. On October 23, a Google team posted a paper to arXiv (2610.32770), \"On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models,\" arguing for the opposite extreme: strip all of that away, hand the model a few coarse topic descriptions, and let it write its own questions, grade its own work, and level itself up.\n\n## Handing the Question-Writer Role to the Model\n\nThe starting point is a well-trodden direction: on-policy self-distillation, which post-trains a model without relying on a stronger external teacher. But as the authors note, existing approaches still lean on ground-truth QA pairs, human-curated problem sets, or other supervised signals — the crutch of human data never fully goes away.\n\nThe framework in this paper (OPST) removes the crutch more thoroughly. The entire pipeline needs exactly one kind of human input: coarse, high-level topic descriptions. Around those topics, the model synthesizes its own questions, its own \"privileged reasoning demonstrations,\" and its own curriculum — from question generation to solving to filtering, without a single human-written prompt or answer.\n\n## Two Gates Against Self-Deception\n\nWhen a model writes its own questions, the obvious risk is mistaking wrong questions for right ones. OPST's answer is a pseudo-verification filter: a self-generated problem only stays in the curriculum if the model can reproduce its answer in a fresh context and under a paraphrase. Reproducible, or it is out.\n\nThe second gate is the privileged reasoning demonstration itself. In the discussion section of the Hugging Face paper page, the authors put the evaluation scope more directly: what they measure is not just in-domain self-improvement, but also whether knowledge is retained on domains that did not participate in self-training when generation domains are mixed. The finding holds on both ends: multi-domain gains in-domain, no loss out-of-domain — a positive signal for the catastrophic-forgetting problem that has long haunted fine-tuning.\n\n## Beyond Text: Visual OPST\n\nGenerality extends past pure text. The paper presents Visual OPST, a multimodal variant that transfers the same self-training logic to visual inputs, achieving consistent self-improvement on Visual Question Answering. The authors' summary in the discussion: OPST improves both language and multimodal reasoning without human-curated questions or answers, and outperforms alternative training strategies.\n\nThe team lists six authors — Leon Sick, Lukas Hoyer, Yannick Strümpler, Christian Reisswig, Timo Ropinski, Federico Tombari — under a Google affiliation. The first author submitted it to Hugging Face Daily Papers on October 26, where it has collected 11 upvotes so far.\n\n## Paper First, Artifacts Pending\n\nOne more fact worth noting: as of writing, the Hugging Face paper page shows zero models, zero datasets, and zero Spaces citing this work. The method is public; the artifacts are not here yet. The paper's own conclusion is measured: foundation models possess rich latent capabilities that can be amplified without external task supervision or curated problem distributions.\n\nThe industry implication cuts two ways. The optimistic read: the closed loop of self-supplied synthetic data takes another step forward — data production shifts from a labor bottleneck to a compute bottleneck, which favors small teams that cannot assemble question banks. The cautious read: the pseudo-verification filter tests consistency, not correctness — a question the model answers confidently and wrongly in every fresh context sails right through. Self-training amplifies the distribution a model already has, which means it amplifies existing biases too, and the paper does not fully close that gap.\n\nA question worth leaving in the comments: when the question-writer, the answerer, and the grader are the same model, how much of your post-training would you hand over to it?","google-opst-data-free-self-training","2026-10-26T21:10:00Z","2026-10-26T21:05:30.067517Z","2026-10-26T21:05:30.067531Z",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},"34edaffc-6b5c-4df1-9e2f-d864cada6063","Gemini 走进 K-12 课堂：Google 把「上下文」塞进每个作业","gemini-classroom-k12-contextualized-prompts","2026-08-07T02:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"23567889-4d40-4af8-a58a-ca61cdb6a8b2","小米A-RSI:不用外部标注,模型自己发现题库刷完了","a-rsi-zero-data-self-improvement","2026-10-30T19:09:58+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"098ef0da-3801-419a-9e33-855d6c6d7ee6","Trinity:递归自我改进,训练基准平均涨 34 分","trinity-capability-driven-recursive-self-improvement","2026-10-28T13:16:03+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"5480e0ca-7b17-481c-8a24-8498f1a0aecd","8B模型OST涨13.41分:空间推理的接口锁定","chronoreplay-spatial-interface-lock-in","2026-10-28T05:09:43+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"f728cc32-7d43-4f75-b4fa-5d76a6115069","全模态幻觉的病根在融合层:OWP 免训练重平衡证据与先验","omni-woodpecker-training-free-hallucination-fix","2026-10-26T05:12:35+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"b95bf7cf-e962-49e8-847c-2e45c917f19d","S-Space:多模态模型看得懂空间,却转不过弯","s-space-multimodal-spatial-workspace","2026-10-12T17:19:44+00:00"]