[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-ncode-llm-automl-searcher":3,"topics-all":38,"news-related-1f38ebab-de2b-4b33-9566-2eaf380eef42":67},{"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},"1f38ebab-de2b-4b33-9566-2eaf380eef42","8B LLM 当 AutoML 搜索器:跨三模态零样本","新论文把 AutoML 重构成 LLM 的排序题：NCode 把架构、超参、数据配比统一编码成数字序列，LLaMA-3.1-8B 只在 CIFAR-10 单域轨迹上训练，就零样本迁移到分割、说话人验证与文生图三模态，论文报告 GenEval 0.91；消融显示直接生成配置会崩 4 个点，排序才是关键。","一个在 CIFAR-10 上测试准确率 97.24% 的神经架构搜索方法 DARTS，换到同样跑 CIFAR-10 的标准测试场 NAS-Bench-201，成绩直接掉到 39.77%。这是新论文《Large Language Models as a Universal Searcher for Open-World Automated Machine Learning》开头讲的故事：今天的 AutoML 工具换个搜索空间就崩，架构、超参、数据配比还要分开调，离「自动」二字尚有距离。这篇论文给出的解法，是训练一个 LLM 来当跨域通用的搜索器。\n\n## 先把所有设计选择变成数字\n\n论文提出的 NCode 是一套统一数值编码：不管你是网络结构（DAG 每条边选什么算子）、训练超参还是数据配比，全部映射成离散数字序列。异构的 AutoML 决策由此落进同一个离散空间，LLM 看到的永远是一串数字加一份历史成绩单。训练数据来自 NAS-Bench-301 的约 10^21 个「架构-性能」对，团队对搜索空间做概率 0.5 的随机边剪枝造出大量子空间，最终攒出约 22.8 万条优化轨迹——全部来自单一视觉分类任务。\n\n## 排序题比作文题好做\n\n方法里最关键的一步是目标重构：不让 LLM 直接生成完整配置，而是给它历史轨迹加一个候选池，让它挑出最有希望的那一个。消融给出了残酷对比——在 ImageNet16-120 上，去掉排序重构、强迫模型直接生成 NCode，测试准确率从 46.51% 崩到 42.53%。主搜索器用 LLaMA-3.1-8B 做全参数微调，上下文窗口设到 20480 token 以装下长历史。也就是说，模型在一域学会的只是「怎么挑」，没有任何目标任务的梯度更新。\n\n## 只学过一个域，三个模态零样本\n\n部署阶段全部零样本：图像分割上搜 SAM 的 LoRA 设计（约 10^19 的架构空间），在医学影像域搜索，配置零样本迁移到伪装物体、阴影、农业叶片和遥感道路数据集，CVC-612 上的 S-measure 比最强基线 Conv-LoRA 绝对高 2.5 个点；音频上做 3D-Speaker 说话人验证，平均 EER 9.21%，优于 ERes2NetV2 的 9.37% 和 OptFormer 的 10.20%；规模最大的一仗是文生图——联合搜索十亿参数级 Diffusion Transformer 的架构、超参与数据配比（论文结论称 1.3B），总计投入 896 GPU 天，论文报告 GenEval 总分 0.91。消融数据同样说明联合搜索的价值：只搜架构得 0.86，只搜超参与数据配比得 0.89，三者联合 0.91。\n\n## GRPO 训出 CoT，把搜索器的脑子剖开\n\n团队还用 GRPO 给 Qwen3-8B 训了自解释思维链，把搜索器的推理归成六种行为模式，并发现清晰的策略转变：早期主要靠邻域变异（在高分配置邻域做小扰动）广泛探索，后期转向锚定利用。策略还会随任务复杂度分化：在简单的 NAS-Bench-201 上，全局组件分布策略的使用率从 10% 涨到近 20%；而在异质的文生图联合搜索任务上，这个比例始终压在 5% 左右——模型「知道」联合优化时不能靠简单数数。\n\n论文地址：https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.12791\n\n## 所以呢\n\n这篇论文值得记的不是某个榜单数字，而是两个判断：其一，LLM 的排序能力比生成能力更可迁移——挑出好答案比从零写出好答案，跨域门槛低得多；其二，AutoML 的下一步未必是更精巧的专家算法，而可能是一个在通用优化轨迹上预训练、到哪都能用的搜索器。当然，896 GPU 天的搜索成本说明「通用」并不便宜，这套方法也还没碰过 LLM 自己的预训练搜索空间。当调参这件事被模型学会，调参工程师的护城河还剩多宽？","https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.12791","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",{"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},"15507678-e47e-44c5-a285-0f80434fe244","en","8B LLM as an AutoML Searcher: Zero-Shot Across Modalities","AutoML reframed as ranking: NCode turns architectures, hyperparameters and data ratios into numeric sequences, and an 8B LLM trained on CIFAR-10 trajectories generalizes zero-shot across vision, audio and text-to-image.","A neural architecture search method that reaches 97.24% test accuracy on CIFAR-10 within its original search space collapses to 39.77% when evaluated on NAS-Bench-201, a benchmark built on the same dataset. That is the opening story of a new paper, \"Large Language Models as a Universal Searcher for Open-World Automated Machine Learning\": today's AutoML tools break when the search space shifts, and they optimize architectures, hyperparameters, and data ratios in isolation. The proposed fix is to train an LLM as a cross-domain universal searcher.\n\n## Turn every design choice into numbers\n\nThe paper introduces NCode, a unified numerical encoding that maps network structures (which operation sits on each DAG edge), training hyperparameters, and data ratios into discrete numeric sequences. Heterogeneous AutoML decisions thus land in one shared discrete space, and the LLM always sees a string of numbers plus a history of scores. Training data comes from NAS-Bench-301, with roughly 10^21 architecture-performance pairs; stochastic edge pruning at probability 0.5 carves out diverse sub-spaces, yielding about 228,000 optimization trajectories — all derived from a single vision classification task.\n\n## Ranking beats generating\n\nThe pivotal move is reframing the objective: instead of asking the LLM to generate a full configuration, it receives the historical trajectory plus a candidate pool and picks the most promising one. The ablation is brutal — on ImageNet16-120, disabling the ranking reformulation and forcing direct NCode generation drops test accuracy from 46.51% to 42.53%. The main searcher is a fully fine-tuned LLaMA-3.1-8B with a 20,480-token context window to hold the long history. What the model learned in one domain is simply \"how to choose\" — no gradient updates on any target task.\n\n## Trained on one domain, zero-shot across three modalities\n\nDeployment is entirely zero-shot. On image segmentation, it searches SAM's LoRA design over an architecture space of about 10^19; searched purely in medical imaging, the configuration transfers zero-shot to camouflaged objects, shadows, leaves, and remote-sensing roads, beating the strongest baseline Conv-LoRA by an absolute 2.5 S-measure points on CVC-612. On audio, it handles speaker verification with 3D-Speaker, reaching an average EER of 9.21% versus 9.37% for ERes2NetV2 and 10.20% for OptFormer. The largest campaign is text-to-image: jointly searching the architecture, hyperparameters, and data ratios of a billion-parameter Diffusion Transformer (the paper's conclusion says 1.3B), at a total cost of 896 GPU-days, with the paper reporting a GenEval overall score of 0.91. The ablation shows why joint search matters: architecture-only reaches 0.86, hyperparameter-and-data-only 0.89, and the joint search 0.91.\n\n## GRPO-trained CoT opens up the searcher's mind\n\nThe team also trains a self-explaining chain-of-thought in Qwen3-8B via GRPO, distilling the searcher's reasoning into six behavioral patterns, and finds a clear strategic shift: early search leans on neighborhood mutation (small perturbations around high-scoring configurations) for broad exploration, then pivots to anchor exploitation. Strategy also adapts to task complexity: on the simpler NAS-Bench-201, the global-component-distribution strategy rises from 10% to nearly 20% usage, while on the heterogeneous text-to-image joint search it stays around 5% — the model \"knows\" that joint optimization cannot rely on simple counting.\n\nPaper: https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.12791\n\n## So what\n\nThe durable lesson is not any single leaderboard number but two judgments. First, an LLM's ranking ability transfers across domains far more easily than its generating ability — picking a good answer from a pool is a lower bar than writing one from scratch. Second, the next step for AutoML may not be ever-more-elegant expert algorithms but a searcher pretrained on generic optimization trajectories that works anywhere. Of course, 896 GPU-days shows that \"universal\" is not cheap, and the method has yet to touch the pretraining search space of LLMs themselves. Once models learn to tune models, how wide is the moat left for the people who tune them?","ncode-llm-automl-searcher","2026-11-10T23:30:00Z","2026-11-10T23:07:59.797850Z","2026-11-10T23:07:59.797860Z",true,"agent",229,[39,49,58],{"slug":40,"tag_slug":41,"title_zh":42,"title_en":43,"intro_zh":44,"intro_en":45,"id":46,"is_active":35,"created_at":47,"modified_at":48},"image-generation","text-to-image","图像生成：从文生图到逐框编辑的模型竞赛","Image Generation: from text-to-image to frame-level editing","文生图已是红海，战场正转向编辑与可控性：FLUX.2\u002F3 用开源和量化把生图搬进消费级显卡，Qwen-Image、Seedream 等国产模型紧咬不放，GPT Image 2 则把推理能力带进像素。本专题追踪图像生成与编辑模型的发布、架构演进、加速量化与商业化数据，看「好图」如何变成「可用的图」。","Text-to-image is a crowded race, and the frontier has moved to editing and control: FLUX.2\u002F3 push open weights and quantization onto consumer GPUs, Qwen-Image and Seedream close in from China, and GPT Image 2 brings reasoning into pixels. This topic tracks model releases, architecture shifts, acceleration work, and the business numbers behind \"pretty\" becoming \"usable\".","52de7c59-7864-4569-8ae7-3d656d26a1a1","2026-11-20T07:20:23.534506Z","2026-11-20T07:20:23.534527Z",{"slug":50,"tag_slug":50,"title_zh":51,"title_en":52,"intro_zh":53,"intro_en":54,"id":55,"is_active":35,"created_at":56,"modified_at":57},"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":59,"tag_slug":59,"title_zh":60,"title_en":61,"intro_zh":62,"intro_en":63,"id":64,"is_active":35,"created_at":65,"modified_at":66},"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":68},[69,74,79,84,89,94],{"id":70,"title":71,"news_slug":72,"published_at":73},"0b895f86-1c24-4c60-9f1f-c2fd0da42b2c","PO降幻觉:提前闭嘴能复现58%到86%的收益","po-hallucination-gain-reproduced-by-early-stopping","2026-11-19T01:30:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"cde1231e-8cfc-428e-9880-2d9a803c2133","同一模型换个读法掉20分:多模态评测的答案读出协议","multimodal-eval-answer-readout-protocol-score-drift","2026-11-17T07:20:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"51e65509-14a8-442b-bc70-9864569dc57f","8个开源VLM数数实测:硬上限大约5个物体","vlm-counting-horizon-geoscope","2026-11-10T09:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"e4ac848d-d24a-465f-bece-37e090da86de","全双工语音模型长对话必崩:KAIST 揪出初始化锚点被挤走","full-duplex-speech-collapse-fd-lhb-kv-cache","2026-10-27T03:20:00+00:00",{"id":90,"title":91,"news_slug":92,"published_at":93},"42f5ea83-365c-4a4e-a318-17d7c4c11c15","论文实测:自洽性证书在AIME上24%的题上会误触发","self-consistency-certificates-misfire-correlated-reasoning","2026-11-21T09:30:00+00:00",{"id":95,"title":96,"news_slug":97,"published_at":98},"53dd30bb-ba19-45b6-a2de-5011709608b6","论文实测:一段文本认出39%审稿人,靠的是曝光量","frontier-models-name-authors-by-exposure","2026-11-21T01:40:00+00:00"]