New adapter embeds foundation models in discrete-choice models, preserving economic logic

Researchers have developed a novel two-stage adapter to integrate foundation models with discrete-choice models, ensuring economic logic is preserved. This method embeds foundation model predictions within a multinomial logit structure, fitting structural coefficients with sign constraints and then applying a neural correction. The approach guarantees the preservation of marginal rates of substitution and has demonstrated significant accuracy gains, improving test accuracy by an average of 6.4 percentage points across various datasets and foundation models, while maintaining cost monotonicity. AI

IMPACT This research offers a method to ensure AI predictions align with economic principles, potentially improving the reliability of AI in economic forecasting and decision-making.

RANK_REASON The cluster describes a new academic paper detailing a novel method for integrating foundation models with discrete-choice models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Discrete choice models with multiple unobserved choice characteristics
• foundation model
• marginal rate of substitution
• Multinomial Logit
• Neural Correction
• transport economics
• value of time

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →