Selective Matching Losses -- Not All Scores Are Created Equal

TL;DR

This paper introduces a new class of selective matching loss functions that emphasize important score regions, improving prediction accuracy in critical areas for various applications like ranking and LLM fine-tuning.

Contribution

It proposes a novel framework for designing selective losses using link functions, extending to multi-class scenarios with composite Softmax, addressing limitations of traditional loss functions.

Findings

Selective losses outperform traditional losses in high-importance regions.

The framework effectively handles multi-class and ranking tasks.

Applications include ranking, retrieval, distillation, and LLM alignment.

Abstract

Learning systems match predicted scores to observations over some domain. Often, it is critical to produce accurate predictions in some subset (or region) of the domain, yet less important to accurately predict in other regions. We construct selective matching loss functions by design of increasing link functions over score domains. A matching loss is an integral over the link. A link defines loss sensitivity as function of the score, emphasizing high slope high sensitivity regions over flat ones. Loss asymmetry drives a model and resolves its underspecification to predict better in high sensitivity regions where it is more important, and to distinguish between high and low importance regions. A large variety of selective scalar losses can be designed with scaled and shifted Sigmoid and hyperbolic sine links. Their properties, however, do not extend to multi-class. Applying them per dimension lacks ranking sensitivity that assigns importance according to class score ranking. Utilizing composite Softmax functions, we develop a framework for multidimensional selective losses. We overcome limitations of the standard Softmax function, that is good for classification, but not for distinction between adjacent scores. Selective losses have substantial advantage over traditional losses in applications with more important score regions, including dwell-time prediction, retrieval, ranking with either pointwise, contrastive pairwise, or listwise losses, distillation problems, and fine-tuning alignment of Large Language Models (LLMs).