[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-limix-2-tabular-foundation-model":3,"topics-all":38,"news-related-8def771a-d936-4859-930d-02c3011dc55c":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},"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","Stable AI 开源 4 亿参数表格基础模型，一次前向完成分类、回归、插补，免微调，TabArena、TALENT、BCCO 三大基准全部第一。","表格数据是企业里最常见的业务形态，却是基础模型浪潮里最被冷落的一块。9 月 16 日，Stable AI 把 LimiX-2 的权重和推理代码开进了 [GitHub 仓库](https:\u002F\u002Fgithub.com\u002Flimix-ldm-ai\u002FLimiX)：4 亿参数的预训练模型，一次前向完成分类、回归、缺失值插补三类任务，不需要针对下游任务更新任何参数。论文挂在 arXiv:2609.17488，当天冲上 Hugging Face Daily Papers 榜首。\n\n## 三大基准全部第一，含金量在对手\n\n官方跑分是这次发布最硬的部分。TabArena 总榜 Elo 1935，领先第二名 TabFM+ 117.4 分，聚合胜场约为对手 3.6 倍；TALENT 1506，领先 35 分，五个类别全部第一；BCCO 1432，对 AutoGluon 1.6 和 TabFM 分别领先 56、63 分。分类子榜 38 个数据集胜率 94.5%，回归子榜 13 个 96.9%。\n\n参照系不是凑数：AutoGluon 1.6 是亚马逊的 AutoML 系统，TabFM、TabPFN-3、EXAONE Tabular 是表格基础模型赛道的主要玩家，CatBoost 是梯度提升阵营的底线。单一预训练模型把这些方法整体压在身后——TabArena 榜上 TabPFN-3 落后三百多分。\n\n## 把\"预测目标列\"换成\"学机制\"\n\n核心改动在建模对象。此前表格基础模型大多围绕 p(y|x, D_context) 组织——给特征、预测目标列，本质是\"指定考题\"式训练。LimiX-2 采用 CMN（Contextual Mechanism Network）范式，转向学习 p(x, y | D_context)：整张表的联合生成结构是学习目标，任何变量都可从其余变量推断。\n\n落到预训练上是 CCMM：同一张表在不同观测模式下遮住不同变量，裂变成大量条件预测问题。训练数据全部来自结构因果模型（SCM）合成，不依赖真实表格语料。附带产物是因果感知——特征注意力编码直接因果关系，可用于因果骨架恢复，超出\"预测\"的范畴。\n\n## 4 亿参数远不是上限\n\nscaling 部分可能是最值得盯的信号。团队评估了 12.5M 到 406.2M 的五条参数系列，下游性能随规模呈清晰的 log-linear 趋势，测量范围内没有饱和迹象。TabArena 上每翻倍参数量 Elo 提升约 34.7 分，R² 达 0.98——曲线还远没走到头，更大的模型仍有稳定收益。\n\n工程侧：推理代码 pip 可装（Python 3.12 + PyTorch 2.9.1），支持 CPU 与 GPU；checkpoint 已放上 Hugging Face，仓库两天攒下 4.2k star、304 fork。\n\n## 该泼的冷水：许可证与自报跑分\n\n两点要冷静。其一，权重采用 StableAI LimiX Non-Commercial License，非商用；代码协议虽以 Apache 2.0 为底，但追加了署名与模型命名条款，上生产前先看清。其二，上述跑分均为官方自报，第三方独立复现还没有出现，\"全部第一\"当前只能以官方口径对待。\n\n意义在路线本身：语言模型跑通\"预训练 + 免微调\"之后，第一次有团队用 scaling law 指引着把同样的范式推到表格三榜第一，并明确展示了未饱和的外推空间。企业里最重的活从来不是写诗，而是表格——这条赛道若继续按 log-linear 兑现，AutoML 四小时调参的默认地位未必保得住。\n\n参考：arXiv:2609.17488 \u002F github.com\u002Flimix-ldm-ai\u002FLimiX \u002F huggingface.co\u002Fstable-ai\u002FLimiX-2","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.17488","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},"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},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"45a66860-ea5e-4462-9c71-cf15fcc8a6db","en","LimiX-2 Open-Sourced: One Tabular Model Tops 3 Benchmarks","Stable AI's 400M tabular model LimiX-2 goes open-weight: one pass covers classification, regression, and imputation, topping all three benchmarks.","Tabular data is the most common data form in enterprises, yet the most neglected by the foundation-model wave. On September 16, Stable AI released the weights and inference code of LimiX-2 to its [GitHub repo](https:\u002F\u002Fgithub.com\u002Flimix-ldm-ai\u002FLimiX): a 400M-parameter pretrained model that handles classification, regression, and missing-value imputation in a single forward pass, with no task-specific parameter updates. The paper landed on [arXiv:2609.17488](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.17488) a day earlier and reached #1 on Hugging Face Daily Papers that day.\n\n## First on all three benchmarks — against real opponents\n\nThe official scores are the hardest part of this release. On TabArena, LimiX-2 scores an overall Elo of 1935, leading runner-up TabFM+ by 117.4 points with an aggregated win count of 18.9 — roughly 3.6x the opponent's. On TALENT it reaches 1506, 35 points ahead, ranking first across all five evaluation categories. On BCCO it hits 1432, leading AutoGluon 1.6 and TabFM by 56 and 63 points respectively. The TabArena classification subset (38 datasets) shows a 94.5% win rate; the regression subset (13 datasets) shows 96.9%.\n\nThe comparison field is not padding: AutoGluon 1.6 is Amazon's AutoML system; TabFM, TabPFN-3, EXAONE Tabular, and Mitra-v2 are the main players of the tabular foundation-model lane; CatBoost and RandomForest are the floor of the gradient-boosting camp. A single 400M pretrained model puts all of them behind — on TabArena, TabPFN-3 trails by over 300 Elo points.\n\n## From \"predict the target column\" to \"learn the mechanism\"\n\nThe core change is what gets modeled. Most tabular foundation models organize around p(y|x, D_context) — given features, predict the target column — which is still \"answering a designated question.\" LimiX-2 adopts the CMN (Contextual Mechanism Network) paradigm proposed in the paper, shifting to learning p(x, y | D_context): the joint generative structure of the whole table, from which any variable can be inferred from the rest.\n\nIn pretraining this becomes CCMM (Context-Conditional Masked Modeling): the same table is masked under different observation patterns, turning one table into many conditional prediction problems — inferring different unknowns from different evidence. Training data comes entirely from synthetic datasets generated by structural causal models (SCMs), spanning diverse graph structures, functional mechanisms, and observation processes, with no real tabular corpora involved. A byproduct is causal awareness: the model's feature attention encodes direct causal relationships and can be used for causal skeleton recovery — beyond prediction, into understanding how variables influence each other.\n\n## 400M is far from the ceiling\n\nThe scaling section may be the signal most worth watching. The team evaluated five parameter series from 12.5M to 406.2M; downstream performance follows a clear log-linear trend with model size, with no sign of saturation within the measured range. On TabArena, the fitted trend gives roughly 34.7 Elo per parameter doubling, with R² at 0.98. The curve has room left — bigger tabular foundation models still buy steady gains.\n\nEngineering openness is solid: inference ships as a pip-installable package (Python 3.12 + PyTorch 2.9.1, CPU and GPU), the LimiX-2 checkpoint is on Hugging Face, and the GitHub repo gathered 4.2k stars and 304 forks within days.\n\n## Cold water: license and self-reported scores\n\nTwo caveats. First, the weights ship under the StableAI LimiX Non-Commercial License. The repo code license is Apache-2.0-based but adds attribution and model-naming provisions. Teams planning production deployment should read the terms first. Second, all scores above are vendor self-reported; no independent third-party reproduction has appeared yet, so \"first on everything\" currently holds only as an official claim. Community comments so far are congratulatory — substantive challenges or verifications have not shown up.\n\nFor the industry, the significance is the route itself: after language models proved \"pretrain + no fine-tuning,\" this is the first time a team has used scaling laws to push the same paradigm to first place across three tabular benchmarks, with an explicit unsaturated extrapolation ahead. The heaviest enterprise workloads were never poem-writing — they were tables. If this lane keeps paying off log-linearly, the default status of AutoML's four-hour tuning sessions may not hold.\n\nRefs: [arXiv:2609.17488](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.17488) \u002F [GitHub: limix-ldm-ai\u002FLimiX](https:\u002F\u002Fgithub.com\u002Flimix-ldm-ai\u002FLimiX) \u002F [HF model page](https:\u002F\u002Fhuggingface.co\u002Fstable-ai\u002FLimiX-2)","limix-2-tabular-foundation-model","2026-09-17T21:09:27Z","2026-09-17T21:09:51.038624Z","2026-09-17T21:09:51.038636Z",true,"agent",44,[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},"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":65,"title":66,"news_slug":67,"published_at":68},"d41175a7-ad10-4e00-9017-a148fa0a77b3","BenchMIRT 把 LLM 基准拆到单题:Ai2 想让模型排名不再「一张考卷定生死」","ai2-benchmirt-llm-benchmark-audit","2026-09-10T11:05:05+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"5bfdf32b-44eb-4eb5-a98b-39e921168182","九天内连发五款前沿模型:7 月的大模型军备赛,真正决胜负的不再是 benchmark","july-2026-five-frontier-models","2026-07-23T12:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"8173a86b-4e5e-429a-8ddf-f98af527b4b5","LLM-as-a-Verifier：验证成 LLM 第四 scaling 维度","llm-as-a-verifier-fourth-scaling","2026-07-07T12:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"ec2c558c-502d-43a5-9494-c766dfd515e9","EurekAgent：把科学发现的瓶颈从「工作流」拽到「环境」，11 美元跑出 26 圆 packing 新 SOTA","eurekagent-environment-engineering-11-usd","2026-06-11T17:56:35+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"af09e362-6537-4b62-bf46-8c8c4ce00982","2026 AI Index报告：开源与闭源LLM差距为何重新拉大？","stanford-ai-index-2026-open-vs-closed-3pct","2026-05-30T04:20:00+00:00"]