LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence
Summary
LimiX-2 is a pretrained foundation model for structured data that uses contextual mechanism networks to achieve #1 on major tabular benchmarks, supporting multiple tasks without task-specific parameter updates.
View Cached Full Text
Cached at: 09/17/26, 02:52 AM
Paper page - LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence
Source: https://huggingface.co/papers/2609.17488
https://huggingface.co/papers/2609.17488#%F0%9F%A7%A9-limix-2-towards-general-structured-data-intelligence🧩LimiX-2: Towards General Structured-Data Intelligence
One pretrained model. Multiple structured-data tasks. #1 across three major tabular benchmarks.
LimiX-2 is a pretrained foundation model for structured data that can support 🎯Classification 📈Regression 🧩Missing-value imputation 🔗Causal skeleton recovery — withno task-specific parameter updatesfor downstream prediction tasks.
https://huggingface.co/papers/2609.17488#%F0%9F%8F%86-1-across-all-three-benchmarks🏆 #1 across all three benchmarks
LimiX-2 takes the top spot on all three broad tabular benchmarks we evaluate: 🏆 TabArena — 1935 Elo 🏆 TALENT — 1506 Elo 🏆 BCCO — 1432 Elo LimiX-2 ranks #1overall across all three, ahead of leading tabular foundation models and strong dataset-specific systems. And the gains continue as we scale the model from12.5M → 406.2Mparameters: performance improves consistently across all five evaluated scaling series, with no clear sign of saturation within the measured range. So how can one pretrained model handle such different structured-data problems?
https://huggingface.co/papers/2609.17488#a-table-is-more-than-a-target-columnA table is more than a target column.
Most tabular foundation models are organized around one question: given ( x ), predict ( y ). LimiX-2 takes a broader view: Can a model learn the relationships among variables in a table, so that different unknown quantities can be inferred from different observed ones? We introduceContextual Mechanism Networks (CMNs), a new paradigm that shifts structured-data in-context learning from target-centric prediction toward context-dependent joint modeling. Instead of focusing only on p(y∣x,Dcontext)p(y \mid x, D_{\mathrm{context}}) CMNs aim to capture the broader dependency structure behind p(x,y∣Dcontext)p(x,y \mid D_{\mathrm{context}}) LimiX-2 instantiates this idea withContext-Conditional Masked Modeling (CCMM). During pretraining, different variables are masked under different observation patterns, turning the same table into many conditional prediction problems. Rather than learning only how to predict one designated target, the model learns to infer different variables from different available evidence. LimiX-2 is pretrained exclusively on synthetic datasets generated from diversestructural causal models (SCMs), spanning different graph structures, functional mechanisms, distributions, and observation processes. This broader training objective gives a single pretrained model a unified basis for classification, regression, and missing-value reconstruction. Beyond predictive performance, we also find that LimiX-2’s learned feature attention contains structural information that can be used for causal skeleton recovery—suggesting that learning conditional relationships across variables can capture structure beyond a single supervised target. This points to a broader possibility:
https://huggingface.co/papers/2609.17488#structured-data-foundation-models-may-move-beyond-predicting-predefined-targets-toward-learning-reusable-mechanisms-of-how-variables-relate-and-interactstructured-data foundation models may move beyond predicting predefined targets toward learning reusable mechanisms of how variables relate and interact.
https://huggingface.co/papers/2609.17488#a-table-is-more-than-a-target-column-prediction-may-be-only-the-beginningA table is more than a target column. Prediction may be only the beginning.
The next question is:
https://huggingface.co/papers/2609.17488#how-far-can-general-structured-data-intelligence-scaleHow far can general structured-data intelligence scale?
https://huggingface.co/papers/2609.17488#limix-2-is-just-the-beginning-follow-the-project-as-we-push-toward-larger-models-broader-capabilities-and-more-general-structured-data-intelligence-%F0%9F%9A%80LimiX-2 is just the beginning. Follow the project as we push toward larger models, broader capabilities, and more general structured-data intelligence. 🚀
📄 Paper —https://arxiv.org/abs/2609.17488 💻 Code —https://github.com/limix-ldm-ai/LimiX 🤗 Model —https://huggingface.co/stable-ai/LimiX-2 Reproductions, independent evaluations, integrations, and feedback are very welcome.
Similar Articles
Xiaomi-TabLDM: A Tabular Foundation Model Technical Report
Introduces Xiaomi-TabLDM, a tabular foundation model that leverages synthetic data and in-context learning for superior prediction accuracy without task-specific fine-tuning, achieving top rankings on multiple benchmarks.
From Residuals to Reasons: LLM-Guided Mechanism Inference from Tabular Data
Introduces Multi-Agent Residual In-Context Learning (MARICL), an agentic framework that uses LLM agents to analyze residuals from a base model on tabular data, hypothesize missing structure, and produce explicit correction terms via textual gradient optimization. Across nine benchmarks, MARICL consistently improves over its base model and demonstrates mechanistic generalization in cell-free protein predictions.
LoGIC: Budgeted Context Construction for Node-Level Graph In-Context Learning with Tabular Foundation Models
LoGIC introduces budgeted context construction for node-level graph in-context learning, optimizing labeled and unlabeled node selection to reduce memory usage and enable scalable inference on large graphs.
@Meituan_LongCat: Introducing LongCat-2.0 1.6T parameters · MoE with ~48B active · 1M context The full model behind Owl Alpha on @OpenRou…
Meituan introduces LongCat-2.0, a 1.6T parameter MoE model with ~48B active parameters and 1M context, featuring novel architectures like LongCat Sparse Attention and Zero-Compute Experts, achieving strong benchmark scores on coding and reasoning tasks.
MiniMaxAI/MiniMax-M3
MiniMax releases M3, a native multimodal model with 1M context and ~428B parameters, using MiniMax Sparse Attention (MSA) for efficient long-context processing, achieving frontier-level coding and agentic performance.