The paper introduces RefineICL, an attention‑gated, feed‑forward‑network‑free framework that refines representations in situ for tabular foundation models. By using support labels to guide episode‑specific updates, the method transfers learned corrections to unlabeled queries without altering model parameters, achieving state‑of‑the‑art performance on AMLB29 and TabArena benchmarks. Experiments and internal interventions demonstrate that intermediate support updates are essential for constructing task‑specific predictors in context.
By Tian Zhou, Beverly Jin, Linxiao Yang, Xue Wang, Wenwei Wang, Bingqing Peng, Mengni Ye, Jinjie Gu, Liang Sun
The paper investigates how Tabular Foundation Models (TFMs) can achieve strong transfer learning by self‑supervised pre‑training on a single real table, rather than large synthetic or real datasets. It finds that a table’s usefulness for downstream tasks is mainly determined by the number of features, not instances, and that fine‑grained column‑level preprocessing improves performance while dataset‑level filtering does not. The authors propose a task‑centric, retrieval‑based view of in‑context generalization, suggesting that effective TFMs identify and aggregate relevant examples from the provided context.
The paper investigates how Tabular Foundation Models (TFMs) can achieve strong transfer learning by self‑supervised pre‑training on a single real table. It finds that a table’s usefulness is largely determined by its number of features rather than instances, and that fine‑grained column‑level preprocessing improves downstream performance while dataset‑level filtering does not. The authors propose that tabular in‑context generalization is primarily retrieval‑based, with models learning to identify and aggregate relevant examples from the provided context.
By Nour Shaheen, Junwei Ma, Alex Labach, Frank Hutter, Valentin Thomas, Anthony L. Caterini
SOMTab is a Set-Order Mamba architecture designed for efficient tabular in-context learning. It separates representation construction from query-conditioned retrieval, using Mamba-based state‑space mixing to build compact row and column representations while retaining attention for final prediction. The model, along with a synthetic prior called DCH‑TailMix, achieves performance comparable to Transformer‑based tabular foundation models but with faster inference and lower GPU memory usage.
By Hao Wang, Siyu Zhang, Wei Ma
The paper investigates how synthetic pretraining priors used in tabular foundation models (TFMs) influence downstream performance. By reconstructing the synthetic data generators of four TFMs and comparing their generated tasks to two popular tabular benchmarks using structural descriptors, the authors measure structural coverage and normalized density. They find that some generators provide broader and denser support for benchmark tasks, and that stronger synthetic-to-benchmark support generally correlates with better model performance.
By He Zhao, Ryan Thompson, Daniel M. Steinberg, Ashfaqur Rahman, Edwin V. Bonilla, Cheng Soon Ong
The paper introduces iAmTime, a time‑series foundation model that uses instruction‑conditioned in‑context learning to adapt to tasks at inference time. iAmTime represents each episode as a structured prompt with semantic tokens that focus on specific time‑series regions, enabling the model to infer task structure from input‑output demonstrations. Trained on large real and synthetic corpora across forecasting, imputation, reconstruction, classification, anomaly detection, and source de‑mixing, iAmTime outperforms strong baselines on zero‑shot probabilistic and point forecasting while matching or exceeding performance on several non‑forecasting tasks.
By Anish Saha, Konstantin Shmakov
Xiaomi-TabLDM is a tabular foundation model that performs classification and regression via in-context learning without task‑specific fine‑tuning. It is pretrained solely on synthetic data from structural causal models, achieving top‑ranked regression results on multiple benchmarks while reducing training and prediction time compared to leading models. The architecture incorporates a three‑stage training strategy, dual‑stream feature grouping, lightweight attention residuals, and sparse mixture‑of‑experts, and it can further improve accuracy through test‑time compute scaling.
By TabLDM Team, Penghui Wang, Wei Liu, Hong Wang, Chengyue Huang, Yuxi Sun, Zirui Wang, Hongming Huang, Quan Wang, Chunxiao Liu, Erli Meng, Bin Wang
arXiv:2608. 10837v1 Announce Type: cross Abstract: The strong performance of foundation models for tabular tasks comes at substantial inference costs.
By Mykhailo Koshil, Matthias Feurer, Katharina Eggensperger
arXiv:2606. 06814v1 Announce Type: cross Abstract: The transformer's emergent ability to perform in-context learning (ICL) has sparked a wide range of studies designed to understand its underlying mechanisms.
By Soo Min Kwon, Alec S. Xu, Can Yaras, Dogyoon Song, Laura Balzano, Qing Qu
arXiv:2609.36337v1 Announce Type: new
Abstract: Tabular foundation models achieve strong performance by conditioning on labelled examples in context, but softmax attention limits their use on large d...
By David Schnurr, Felix Sarnthein, Thomas Hofmann, Imanol Schlag
The paper introduces SIMPLE, a prior‑fitted multi‑view in‑context learner that learns a reusable, task‑conditioned inference procedure instead of a fixed fusion function. By generating synthetic task priors in embedding space, SIMPLE can handle diverse view configurations, class structures, and missingness patterns. Experiments on multi‑view and multi‑omics benchmarks show that a frozen SIMPLE model performs competitively, and lightweight adapter calibration further improves performance across most datasets.
By Jielong Lu, Zhihao Wu, Jiajun Yu, Zhaoliang Chen, Haishuai Wang
Transformers can learn broad families of tasks during pretraining and adapt to unseen tasks from a short prompt, but a rigorous understanding of this capability is limited. This paper studies how shared cross‑task structure influences the sample complexity of in‑context learning (ICL) by characterizing task‑space complexity through covering numbers, yielding a set of anchor functions that localize unseen tasks and predict responses. The authors construct a Transformer with Softmax attention to approximate this procedure and derive an error bound that separates the effects of pretraining tasks and prompt length, showing that once enough tasks are available the dependence on prompt length becomes dimension‑free.
By Zhongjie Shi, Rongjie Lai, Alexander Cloninger, Wenjing Liao