Post Hoc Regression Refinement via Pairwise Rankings

Minghao Guo (MIT) · Wojciech Matusik (MIT) · Kevin Tirta Wijaya (Max Planck Institute for Informatics) · Michael Sun (Computer Science and Artificial Intelligence Laboratory, Electrical Engineering & Computer Science) · Hans-peter Seidel (Max-Planck Institute) · Vahid Babaei (Max Planck Institute for Informatics)

data-scarce regimesdeep-learning regressorsexpert knowledge injectiongeneral-purpose llmsinverse-variance weightinglarge language model low-data settings mean absolute error model-agnostic molecular property predictionpairwise rankingspost-hoc refinementpractical applicabilityrank-based estimaterankrefineregression improvement

Accurate prediction of continuous properties is essential to many scientific and engineering tasks. Although deep-learning regressors excel with abundant labels, their accuracy deteriorates in data-scarce regimes. We introduce RankRefine, a model-agnostic, plug-and-play post-hoc refinement technique that injects expert knowledge through pairwise rankings. Given a query item and a small reference set with known properties, RankRefine combines the base regressor’s output with a rank-based estimate via inverse-variance weighting, requiring no retraining. In molecular property prediction task, RankRefine achieves up to 10\% relative reduction in mean absolute error using only 20 pairwise comparisons obtained through a general-purpose large language model (LLM) with no finetuning. As rankings provided by human experts or general-purpose LLMs are sufficient for improving regression across diverse domains, RankRefine offers practicality and broad applicability, especially in low-data settings.