Interpretable Tabular Foundation Models via In-Context Kernel Regression
This work addresses the problem of interpretability in tabular foundation models for users needing transparent predictions, though it is incremental as it modifies existing architectures without fundamental changes.
The paper tackled the opacity of tabular foundation models by introducing KernelICL, a framework that enhances interpretability through explicit kernel regression mechanisms, achieving performance on par with existing models on 55 benchmark datasets.
Tabular foundation models like TabPFN and TabICL achieve state-of-the-art performance through in-context learning, yet their architectures remain fundamentally opaque. We introduce KernelICL, a framework to enhance tabular foundation models with quantifiable sample-based interpretability. Building on the insight that in-context learning is akin to kernel regression, we make this mechanism explicit by replacing the final prediction layer with kernel functions (Gaussian, dot-product, kNN) so that every prediction is a transparent weighted average of training labels. We introduce a two-dimensional taxonomy that formally unifies standard kernel methods, modern neighbor-based approaches, and attention mechanisms under a single framework, and quantify inspectability via the perplexity of the weight distribution over training samples. On 55 TALENT benchmark datasets, KernelICL achieves performance on par with existing tabular foundation models, demonstrating that explicit kernel constraints on the final layer enable inspectable predictions without sacrificing performance.