Real-Time Explanations for Tabular Foundation Models
This work addresses the need for real-time interpretability in scientific machine learning, enabling interactive exploration for researchers and practitioners, though it is incremental as it builds on existing foundation models and Shapley value methods.
The paper tackles the problem of slow explanation methods for tabular foundation models by introducing ShapPFN, which integrates Shapley value regression into its architecture to produce predictions and explanations in a single forward pass, achieving competitive performance with high-fidelity explanations (R^2=0.96, cosine=0.99) over 1000 times faster than KernelSHAP (0.06s vs 610s).
Interpretability is central for scientific machine learning, as understanding \emph{why} models make predictions enables hypothesis generation and validation. While tabular foundation models show strong performance, existing explanation methods like SHAP are computationally expensive, limiting interactive exploration. We introduce ShapPFN, a foundation model that integrates Shapley value regression directly into its architecture, producing both predictions and explanations in a single forward pass. On standard benchmarks, ShapPFN achieves competitive performance while producing high-fidelity explanations ($R^2$=0.96, cosine=0.99) over 1000\times faster than KernelSHAP (0.06s vs 610s). Our code is available at https://github.com/kunumi/ShapPFN