arXiv · 2608.06137
SkillTFM: Gated Skill Evolution for Training-Free Adaptation of Tabular Foundation Models
Abstract
Tabular data are ubiquitous in real-world applications and are crucial for data-driven prediction and decision-making across science, industry, finance, healthcare, and public services. Tabular foundation models (TFMs) have emerged as a promising paradigm for general-purpose tabular learning, offering reusable predictors across diverse datasets and substantially reducing the need for task-specific training, tuning, and model development. However, their practical deployment remains constrained by distribution shifts, heterogeneous feature semantics, and task-specific patterns that are difficult to capture without costly fine-tuning or additional labeled data. To this end, we propose SkillTFM, a training-free system that shifts TFM adaptation from parameter updates to the gated evolution of agentic skills. The core of SkillTFM is a verifiable and extensible skill bank that couples boundary evidence identification with gated skill evolution: the former characterizes task structure and base-model failure patterns, whereas the latter retrieves and extends reusable skills subject to explicit validation. Across simulated boundary settings and real-world electricity-price forecasting, SkillTFM improves AUC by 0.128--0.142, raises nonlinear-boundary AUC from 0.699 to 0.898. Furthermore, experiments across TFM backbones demonstrate the effectiveness and generality of SkillTFM.
Explore related subjects
Keep this discovery
Yi He, Zhengkang Guan, Anpeng Wu, Peng Cui, Fei Wu, Kun Kuang. 2026-08-06. SkillTFM: Gated Skill Evolution for Training-Free Adaptation of Tabular Foundation Models. https://arxiv.org/abs/2608.06137
Cite the original work for its findings. Save a collection to share your selection of sources.