Conformal Candidate Certification for Offline Model-Based Optimization

2026-06-13

AI Summary

This work addresses the unreliability of surrogate model scores in offline model optimization due to out-of-distribution extrapolation and the lack of statistical guarantees for satisfying target thresholds. To tackle these issues, the authors propose a posterior calibration method that integrates entropy-regularized surrogate maximization with weighted conformal prediction. Their approach leverages the same surrogate model to directly generate importance weights, thereby avoiding separate density ratio estimation and enabling reliable one-sided lower-bound certification under distribution shift. Empirical results demonstrate that the method achieves 99.0% actual coverage at a nominal 90% confidence level and successfully certifies 16.7% of aggressive candidate solutions, whereas baseline methods ignoring covariate shift suffer a drastic drop in coverage to 41.6%.

Abstract

Offline model-based optimization (MBO) proposes candidates by optimizing a surrogate trained on a fixed historical dataset. Because candidates are deliberately out-of-distribution, surrogate rankings are least reliable exactly where the optimizer is most aggressive, yet existing methods provide no per-candidate statistical certificate that a design meets a target threshold. We propose \emph{Conformal Candidate Certification} (CCC), a post-hoc wrapper that attaches a calibrated one-sided lower bound to each candidate and advances only those whose bound exceeds the target. We show that entropy-regularized surrogate maximization induces a Gibbs-tilted proposal, so the same surrogate supplies importance weights for weighted conformal prediction without a separate density-ratio estimation step. In a controlled synthetic study, CCC certifies $16.7\%$ of an aggressive proposal pool with empirical coverage 0.990 at nominal 0.90, while standard conformal prediction ignoring the covariate shift collapses to 0.416 coverage.

Problem

• offline model-based optimization
• out-of-distribution
• statistical certification
• conformal prediction
• surrogate reliability

Innovation

• Conformal Prediction
• Offline Model-Based Optimization
• Covariate Shift
• Importance Weighting
• Entropy Regularization