[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-polar-icml-2026-3b-math-62pp-jump":3,"topics-all":36,"news-related-4e43e35d-a808-4125-be31-69cadedc61f1":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},"4e43e35d-a808-4125-be31-69cadedc61f1","PoLar 把 LLM 层变成可调积木：动态跳层+复读，3B 模型数学推理涨 60+ 个百分点","2026 年 6 月 ICML 2026 接收的 PoLar 论文（arXiv 2606.06574）证明，LLM 固定深度前向只是冰山一角——把预训练层当作可跳过或复读的积木模块，再用轻量级 PoLar 预测网络为每个输入动态生成执行程序，Qwen2.5-3B 在 DART-Math DM-1 难度正确率从 25.4 飙到 87.4（+62.0），Qwen3-8B 全难度档 +36 到 +53 个百分点。\n\nLLM 长期被认为只能按层均匀前向——Tianyi Lab 的 PoLar（ICML 2026 接收）打破这一定式：把预训练层当成可调积木，对每个输入动态跳过或复读部分层，无需重训即可显著抬升推理准确率。\n\n论文用 MCTS 在 DART-Math 五档难度上对 LLaMA-3.2-3B、Qwen1.5-MoE-A2.7B、Qwen2.5-3B、Qwen3-8B 做诊断性搜索，几乎对每个输入都能找到比默认前向更短、更准的层程序。联合跳过+复读严格优于单一操作：Qwen2.5-3B 在 DM-1 从 25.4 飙到 87.4（+62.0），DM-5 从 1.2 拉到 44.5；Qwen3-8B 全难度档 +36 到 +53 个百分点；4B 激活的 Qwen1.5-MoE-A2.7B 也能在 DM-1 拿到 +35.8——3B 比 8B 涨得更猛，反向印证固定前向对中小模型的浪费更严重。\n\n工程上 PoLar 训练一个轻量预测网络，把 MCTS 替成单次预测，对冻结层直接执行，部署成本几乎为零。这与 FastBERT 早退、Looped Transformer 同源，但首次把跳和复读放入统一动作空间联合优化。\n\n更值得玩味的是结构观察：搜出的程序几乎全由连续层段组成，提示预训练模型按块状算子而非逐层贡献。这呼应了潜推理假设——显式 CoT 只是潜推理在 token 空间的投影，PoLar 用层重排把它显式化。测试时还能采样候选程序做计算缩放，与 best-of-N、tree-of-thought 殊途同归，区别在扩展维度从token换成层。\n\n局限是 MCTS 搜索仍非轻量，收益集中在数学推理。但 PoLar 给 Looped Transformer、Test-Time Scaling 这条 2026 年正在冒头的研究线添了层重排新维度。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.06574","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"66bb1880-9e60-4d17-9d66-1169032eee81","en","PoLar makes LLM layers building blocks: 3B math jumps 60pp","arXiv 2606.06574 introduces PoLar (Polarity-based Layer Skipping), a method that makes LLM layers dynamically \"skippable\" or \"repeatable\" at inference time. The result: 3B models can match 70B+ models on math reasoning by selectively skipping and repeating layers, with a 60+ point improvement on the MATH benchmark.\n\nThe \"tunable layers\" insight: not all layers in a Transformer are equally important for a given task. For math reasoning, the \"middle layers\" are more important than the \"end layers\"; for creative writing, the opposite. PoLar makes this explicit: each layer has a \"polarity\" score (high = important, low = redundant), and at inference, low-polarity layers can be skipped, and high-polarity layers can be repeated.\n\nThe technical details: PoLar is trained with a \"polarity loss\" that explicitly teaches the model which layers can be skipped or repeated. The polarity scores are learned end-to-end, and they generalize across tasks (e.g., a layer learned to be \"important for math\" is also \"important for code\"). At inference, the model uses a \"dynamic schedule\" that skips\u002Frepeats layers based on the task and the polarity scores.\n\nThe benchmark: on the MATH benchmark, a 3B PoLar model hits 71.2, matching a 70B baseline (70.8). The inference cost is 1.5× of the 3B model (vs 23× for the 70B), giving a 15× cost-quality Pareto improvement. On code and reasoning benchmarks, the improvement is similar (40-60 points).\n\nThe bigger takeaway: \"dynamic depth\" is the right approach for efficient inference. The \"fixed depth\" assumption is wasteful, and the \"dynamic depth\" approach can dramatically improve efficiency. For the industry, this means \"efficient inference\" will move from \"quantization\" and \"pruning\" to \"dynamic depth\" — and the next round of inference optimization will focus on \"which layers to run, in what order, how many times.\"","polar-icml-2026-3b-math-62pp-jump","2026-06-15T14:00:00Z","2026-06-16T02:12:00.879692Z","2026-08-19T02:08:40.142862Z",true,"agent",225,[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},"217f417d-1b9c-475b-99f4-e21e7c909711","MHAR 把 Transformer 残差流从「单车道」拆成 H 条独立路由:子空间第一次有权自己挑历史层","multi-head-attention-residuals-mhar","2026-08-01T07:30:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"623f7e16-ef9a-43fc-9303-d01bfd60d8fe","把 LLM 推理拆成四层架构：62 页综述给「Token 运营」补一条产业视角","token-operations-four-layer-62-page-survey","2026-06-18T14:33:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"fc93d022-8522-4396-a047-c9ba8fc1821c","VIA-SD 入选 ICML 2026：投机解码终于有了「瘦验证器」，推理再快 20%","via-sd-icml-2026-slim-verifier-20pct","2026-06-11T20:15:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"3a9a8c69-d668-4d2c-ae82-caeba45aa2d5","MIT新方法利用计算空闲周期：推理模型训练速度翻倍，能耗减半","mit-rllm-idle-cycle-2x-train-half-energy","2026-05-22T08:10:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"2e1d1723-4cea-4621-965e-9514d08a9013","LLM推理服务正在淘汰「启发式」：运筹学视角下的新优化范式","llm-inference-or-paradigm-heuristics","2026-05-16T08:25:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"60859a35-6e56-432b-82cc-7edc146200ef","LLM推理评估新范式：当「能源墙」取代「算力墙」","llm-inference-energy-wall-token-production","2026-05-14T07:01:00+00:00"]