Auditing and Fixing Economic Validity in Tabular Foundation Models for Discrete Choice

2026-05-26

AI Summary

This study addresses the frequent violation of economic rationality—such as demand increasing with price or negative willingness-to-pay—by existing tabular foundation models in discrete choice prediction. To resolve this, the authors propose a two-stage adapter framework: first, estimating parameters of a theoretically grounded structural choice model (e.g., Logit) under utility maximization constraints; then, freezing these parameters and integrating information from the foundation model for refinement. This approach is the first to simultaneously achieve high predictive accuracy and strict adherence to economic consistency, including price-demand monotonicity, while enabling analytical computation of trade-off metrics. Evaluated on two transportation datasets, the method improves prediction accuracy by up to 13 percentage points over standard Logit models and attains 100% economic consistency, substantially outperforming both the original foundation model and conventional distillation techniques.

Abstract

Tabular foundation models achieve strong accuracy on choice prediction tasks, but their predictions often violate the economic logic those tasks require: raising a price sometimes increases predicted demand, and implied willingness-to-pay estimates are frequently negative or implausible. We propose a two-stage adapter that embeds foundation model predictions within a utility-maximization framework. In the first stage, we estimate a standard choice model whose parameters are constrained to obey economic theory. In the second stage, we freeze those parameters and train a correction term that incorporates the foundation model's predictions as additional information. The result is a model that inherits the foundation model's accuracy gains while guaranteeing monotonic price-demand relationships under policy perturbation and producing analytically computable trade-off measures. On two transportation datasets, the adapter recovers up to 13 percentage points of accuracy over a standard logit model while maintaining perfect economic consistency, something neither the raw foundation models nor conventional distillation achieve.

Problem

• economic validity
• tabular foundation models
• discrete choice
• price-demand monotonicity
• willingness-to-pay

Innovation

• tabular foundation models
• economic validity
• discrete choice
• utility maximization
• adapter framework