🤖 AI Summary
This work addresses the computational inefficiency of traditional Shapley value estimation in the presence of feature dependencies, which typically requires numerous conditional expectation evaluations and is ill-suited for acceleration via deep learning. The authors propose the first integration of tabular foundation models—such as TabPFN—into conditional Shapley value estimation, leveraging their in-context learning capabilities to efficiently approximate conditional expectations without retraining. By circumventing conventional Monte Carlo integration or repeated model training strategies, the method achieves substantial gains in computational efficiency. Empirical results across multiple synthetic and real-world datasets demonstrate that TabPFN and its variants consistently attain state-of-the-art or near-optimal explanation quality while requiring only a fraction of the runtime of existing approaches.
📝 Abstract
Shapley values have become a cornerstone of explainable AI, but they are computationally expensive to use, especially when features are dependent. Evaluating them requires approximating a large number of conditional expectations, either via Monte Carlo integration or regression. Until recently it has not been possible to fully exploit deep learning for the regression approach, because retraining for each conditional expectation takes too long. Tabular foundation models such as TabPFN overcome this computational hurdle by leveraging in-context learning, so each conditional expectation can be approximated without any re-training. In this paper, we compute Shapley values with multiple variants of TabPFN and compare their performance with state-of-the-art methods on both simulated and real datasets. In most cases, TabPFN yields the best performance; where it does not, it is only marginally worse than the best method, at a fraction of the runtime. We discuss further improvements and how tabular foundation models can be better adapted specifically for conditional Shapley value estimation.