Tabular foundation models have mostly just been predicting a single target variable. A new paradigm learns the relationships across an entire table instead. Title: LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence URL: 📝 Overview Contextual Mechanism Networks (CMN) learn the full joint distribution p(x, y) across a table, not just the target y, letting a single model handle classification, regression, missing-value imputation, and causal discovery. ❗ Problem it solves Existing PFN-based methods specialize in conditional prediction of the target only, without explicitly modeling the dependency structure between variables. ⚙️ Methodology Cell-level embeddings feed a 24-layer dual-axis transformer, trained via "Context-Conditional Masked Modeling" that masks part of the query rows and jointly predicts both features and targets. It's pretrained at scale on synthetic data generated from structural causal models. 📊 Results On TabArena (51 datasets), it hits an Elo of 1935 vs. 1818 for the baseline; on BCCO (156 datasets), cumulative wins are 2.06x higher. It also took 1st place on F1 score for causal skeleton recovery on all 6 tested datasets, all while using 4x fewer parameters than TabFM. 🔬 Use cases Since the same model handles classification, regression, and causal discovery, it could replace much of a data analysis pipeline with a single foundation model. #TabularData# #MachineLearning#

How do you actually build evaluation benchmarks for AI agents at scale? LangChain just shared their full approach. Title: How We Build Agent Environments & Tasks URL: ❓ What exactly makes up an agent "task"? 💡 A task has three components: an input, an environment, and a test script. The environment hosts agent execution, and a rubric defines scoring criteria. A "world spec" consolidates shared domain knowledge across related tasks — API schemas, data generation methods, trace parsing scripts — in one place. ❓ How do you create tasks efficiently at scale? 💡 LangChain uses a two-step pipeline. First, a "spec generation" phase where a coding agent scans repositories, groups traces, maps credentials, and auto-generates an initial world spec. Then a "Spec2Task" phase converts that spec into runnable evaluation tasks. The spec from the very first task becomes the foundation, iteratively refined through subsequent task creation cycles. ❓ What are the most common mistakes when building evaluation tasks? 💡 Three pitfalls stand out: ・Don't skip running tasks with real agents — paper evaluation won't surface environment flaws ・Calibrate difficulty across model tiers (e.g. gpt-5.6-Luna vs Sol) — what's hard for one may be trivial for another ・Match the data generation method to the data type: LLM-based approaches for free-text, SQL scripts for tabular data ❓ Is a benchmark "done" once you've built it? 💡 Not at all. Continuous improvement from production data is the core idea. Real production traces feed back into cost modeling, prompt simplification validation, and tool configuration testing — the benchmark evolves alongside the system it measures. Treating evaluation environment engineering as ongoing rather than a one-time project is the key practical insight here. #AIAgents# #LLMEvaluation#