[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-ttpo-test-time-policy-optimization":3,"topics-all":35,"news-related-ebbe26a1-f5e6-4aeb-b5be-3c2cbddbd523":54},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":21,"news_slug":28,"published_at":29,"created_at":30,"modified_at":31,"is_published":32,"publish_type":33,"image_url":14,"view_count":34},"ebbe26a1-f5e6-4aeb-b5be-3c2cbddbd523","浙大阿里开源TTPO:免标签测试时训练,Qwen3-1.7B数学均分38.0→45.2","浙大ZJU-REAL实验室与阿里提出TTPO:测试时对无标签题目自我优化,同意多数票的rollout走OPSD蒸馏,不同意的用分组RL只罚自信错误。免标签追平有标签OPSD,Qwen3-1.7B数学TTT平均分从38.0%升到45.2%,非思考模式最高+36.4分,代码已开源。","考试的时候没有标准答案,还能不能边考边学?测试时训练(Test-Time Training,TTT)想做的就是这个:模型拿到一批无标签的测试题,自己采样多个解答,用多数票造出\"伪标签\",再在推理阶段就地更新参数。坑也在这里——多数票一旦投错,错误的\"老师\"会带偏每一个 token。\n\n浙江大学 ZJU-REAL 实验室与阿里巴巴的 11 人团队 8 月 27 日向 arXiv 提交了 TTPO(Test-Time Policy Optimization),次日开源代码,并登上 Hugging Face Daily Papers 热榜。\n\n## 标签缺口,和一个不对称的观察\n\n现有后训练主力——强化学习(RL)和在线策略自蒸馏(OPSD)——在数学推理上进步飞快,但都依赖真值标签,这让它们在测试时训练里直接失效。用多数票伪标签替代真值是自然想到的方案,却很脆:一次错误投票会污染 teacher,误导所有 token。\n\nTTPO 论文给出的关键观察是,这个失败模式是不对称的:**与伪标签不一致的 rollout,无论投票本身对错,通常自己就是错的**。也就是说,不需要知道真值,也能把\"少数派\"判定为高置信错误。\n\n## 一个目标,两条支路\n\n顺着这个观察,TTPO 设计了非对称目标:\n\n- **同意伪标签的 rollout**:走 OPSD 蒸馏,把多数派行为蒸馏回模型;\n- **不同意伪标签的 rollout**:走分组 RL,只惩罚其中的\"自信错误\"。\n\ntoken 级选择进一步细化两条支路:蒸馏分支给已经收敛的位置降权,RL 分支只盯高置信错误。作者强调,即使伪标签频繁出错,两个方向的更新依然有据可依;而模型变强之后,多数票路由本身会给出越来越紧的自监督信号。\n\n## 数字:免标签追平有标签\n\n在 AIME25\u002F26、HMMT25\u002F26、BRUMO25 五个竞赛级基准上,作者报告的结果:\n\n- OpenThoughts 训练后,免标签 TTPO 与用真值训练的 OPSD 基本打平:Qwen3-1.7B 平均 40.1 vs 39.7,4B 58.6 vs 58.4,8B 62.6 vs 61.7;\n- 直接在无标签测试题上做 TTT:Qwen3-1.7B 平均分从 38.0% 提到 45.2%,高于 TTRL(40.2)和 OPSD-TTT(41.9);8B 从 60.7 提到 65.3;\n- 非思考模式(non-thinking)收益最大:1.7B\u002F4B\u002F8B 分别 +25.2\u002F+30.6\u002F+36.4 分,8B 非思考平均到 56.7(其思考模式基线为 58.6)。\n\n工程复现门槛不高:官方环境是 Python 3.10、PyTorch 2.8.0、vLLM 0.11.0,四卡 GPU 加 LoRA,仓库直接给了 1.7B\u002F4B\u002F8B 三档训练脚本和评测入口。\n\n## 边界与\"所以呢\"\n\n该留的界限也说清楚:实验主体集中在竞赛数学;摘要声称\"强跨任务泛化\",但这是作者自报数据,且对比方法(如 TTRL)本身仍在快速演进;伪标签路由在弱模型上的稳定性,依旧取决于多数票质量。\n\n但方向是实打实的:当标注贵到不可得、题目分布又在漂移时,TTPO 把\"模型自己教自己\"从口号变成了带配方、带代码、可复现的工程选项。论文见 [arXiv:2608.27448](https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.27448),代码在 [ZJU-REAL\u002FTTPO](https:\u002F\u002Fgithub.com\u002FZJU-REAL\u002FTTPO)。\n\n如果训练不再需要答案,模型的下限就不再由标注预算决定——这大概是测试时训练这条线最值得盯住的一句话。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.27448","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18],{"id":12,"name":13,"slug":13,"description":14,"color":14},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",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},"15523c78-84d3-4431-9782-2f271ce3dff6","推理",[22],{"id":23,"lang":24,"title":25,"summary":26,"content":27},"f7febc9c-ad55-46b3-b59b-a9e492139f99","en","TTPO From ZJU and Alibaba: Label-Free Test-Time Training Lifts Qwen3-1.7B Math From 38.0 to 45.2","ZJU-REAL lab and Alibaba propose TTPO: models self-optimize on unlabeled test problems at inference time. Rollouts agreeing with the majority vote are distilled via OPSD; disagreeing ones get grouped RL that only penalizes confident errors. Label-free TTPO matches label-supervised OPSD, raising Qwen3-1.7B math TTT average from 38.0% to 45.2%, with up to +36.4 points in non-thinking mode. Code is open-sourced.","Can a model keep learning while it is being examined — with no answer key? That is the promise of Test-Time Training (TTT): given a batch of unlabeled test problems, the model samples multiple solutions, builds pseudo-labels from majority voting, and updates its weights in place during inference. The trap sits exactly there: one wrong vote corrupts the teacher and misleads every token.\n\nAn 11-author team from the ZJU-REAL lab at Zhejiang University and Alibaba Group submitted TTPO (Test-Time Policy Optimization) to arXiv on August 27, released the code the next day, and landed on the Hugging Face Daily Papers trending list.\n\n## The label gap, and one asymmetric observation\n\nThe workhorse post-training methods — Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD) — have driven rapid progress in mathematical reasoning, but both rely on ground-truth labels, which rules them out for test-time training. Swapping ground truth for majority-vote pseudo-labels is the natural fix, yet it is fragile: a single incorrect vote corrupts the teacher and misleads every token.\n\nThe key observation in the TTPO paper is that this failure mode is asymmetric: **rollouts that disagree with the pseudo-label are typically wrong regardless of whether the vote itself is correct**. In other words, you can flag the minority as high-confidence errors without ever knowing the truth.\n\n## One objective, two branches\n\nFollowing that observation, TTPO builds an asymmetric objective:\n\n- **Rollouts agreeing with the pseudo-label** are distilled via OPSD, folding majority behavior back into the model;\n- **Rollouts disagreeing with it** go through Grouped RL that penalizes only the confident errors.\n\nToken-level selection refines both branches: distillation down-weights positions that have already converged, while the RL branch only targets high-confidence mistakes. The authors stress that both updates stay well-grounded even under frequent pseudo-label errors, and that majority-vote routing yields tighter self-supervision as the model improves.\n\n## The numbers: label-free matches label-supervised\n\nAcross five competition-level benchmarks (AIME25\u002F26, HMMT25\u002F26, BRUMO25), the reported results:\n\n- After OpenThoughts training, label-free TTPO roughly matches OPSD trained with ground truth: Qwen3-1.7B averages 40.1 vs 39.7, 4B 58.6 vs 58.4, 8B 62.6 vs 61.7;\n- Training directly on unlabeled test problems (TTT): Qwen3-1.7B average rises from 38.0% to 45.2%, above TTRL (40.2) and OPSD-TTT (41.9); the 8B goes from 60.7 to 65.3;\n- Non-thinking evaluation gains are the largest: +25.2 \u002F +30.6 \u002F +36.4 points for 1.7B \u002F 4B \u002F 8B respectively, with the 8B non-thinking average reaching 56.7 (its thinking-mode base is 58.6).\n\nThe engineering bar is low: the official environment is Python 3.10, PyTorch 2.8.0 and vLLM 0.11.0, on four GPUs with LoRA, and the repo ships training scripts for the 1.7B\u002F4B\u002F8B tiers plus an evaluation entry point.\n\n## Boundaries, and the \"so what\"\n\nCaveats worth stating: the experiments center on competition math; the claimed \"strong cross-task generalization\" is author-reported, and baselines like TTRL are themselves evolving fast; the stability of pseudo-label routing on weaker models still hinges on majority-vote quality.\n\nBut the direction is real. When annotation is unaffordable and the task distribution keeps drifting, TTPO turns \"the model teaches itself\" from a slogan into a documented, coded, reproducible engineering option. The paper is at [arXiv:2608.27448](https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.27448); the code lives at [ZJU-REAL\u002FTTPO](https:\u002F\u002Fgithub.com\u002FZJU-REAL\u002FTTPO).\n\nIf training no longer needs answers, the model's floor is no longer set by the labeling budget — arguably the one line in this test-time training thread most worth watching.","ttpo-test-time-policy-optimization","2026-08-30T15:20:00Z","2026-08-30T15:20:03.589444Z","2026-08-30T15:20:03.589456Z",true,"agent",115,[36,45],{"slug":37,"tag_slug":37,"title_zh":38,"title_en":39,"intro_zh":40,"intro_en":41,"id":42,"is_active":32,"created_at":43,"modified_at":44},"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":46,"tag_slug":46,"title_zh":47,"title_en":48,"intro_zh":49,"intro_en":50,"id":51,"is_active":32,"created_at":52,"modified_at":53},"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":55},[56,61,66,71,76,81],{"id":57,"title":58,"news_slug":59,"published_at":60},"5a274662-0c3f-492c-a0e8-a46c5a0be783","别再自己给自己打分了:Co-RL 让模型互相判卷,无标签 RL 追平有监督","co-rl-peer-reward-label-free-rl","2026-08-20T19:10:00+00:00",{"id":62,"title":63,"news_slug":64,"published_at":65},"88105269-9641-44c3-a705-1cf07314614f","LLM 思维链能看出\"用了几分力\":SARE 给每一步推理做 CT 扫描","step-aware-reasoning-energy-llm-cot","2026-08-04T04:00:00+00:00",{"id":67,"title":68,"news_slug":69,"published_at":70},"0138c087-f1c3-44c5-a662-045c6f20a756","ACL 2026 新基准:大模型的古诗美学推理翻车了","neo-classic-poetry-reasoning-benchmark","2026-09-20T15:15:00+00:00",{"id":72,"title":73,"news_slug":74,"published_at":75},"8def771a-d936-4859-930d-02c3011dc55c","LimiX-2 开源：一个模型吃下分类回归插补，表格三榜登顶","limix-2-tabular-foundation-model","2026-09-17T21:09:27+00:00",{"id":77,"title":78,"news_slug":79,"published_at":80},"176b4807-da61-479f-a514-9381cd13319e","SP3O:3 个锚点修复 PPO critic 的平坦化","sp3o-sparse-critic-supervision","2026-09-17T17:10:01+00:00",{"id":82,"title":83,"news_slug":84,"published_at":85},"2e27016d-b90e-45c7-825a-41fd1e435c80","JHU 新研究:组合持续学习机制,百任务记忆留存从 1.2% 提到 34.9%","compose-cl-long-horizon-memorization","2026-09-16T15:10:00+00:00"]