Extending Minimal Pairs with Ordinal Surprisal Curves and Entropy Across Applied Domains

TL;DR

This paper extends the minimal pairs paradigm by using surprisal curves and entropy to evaluate language models across multiple domains, providing richer insights into model uncertainty and response preferences beyond binary judgments.

Contribution

It introduces a novel surprisal-based evaluation framework that applies ordinal-scaled scoring and measures model uncertainty across diverse tasks, moving beyond binary grammaticality judgments.

Findings

Surprisal curves produce interpretable classification signals.

Entropy distinguishes ambiguous items from easier ones.

Framework applicable across multiple domains.

Abstract

The minimal pairs paradigm of comparing model probabilities for contrasting completions has proven useful for evaluating linguistic knowledge in language models, yet its application has largely been confined to binary grammaticality judgments over syntactic phenomena. Additionally, standard prompting-based evaluation requires expensive text generation, may elicit post-hoc rationalizations rather than model judgments, and discards information about model uncertainty. We address both limitations by extending surprisal-based evaluation from binary grammaticality contrasts to ordinal-scaled classification and scoring tasks across multiple domains. Rather than asking models to generate answers, we measure the information-theoretic "surprise" (negative log probability) they assign to each position on rating scales (e.g., 1-5 or 1-9), yielding full surprisal curves that reveal both the model's preferred response and its uncertainty via entropy. We explore this framework across four domains: social-ecological-technological systems classification, causal statement identification (binary and scaled), figurative language detection, and deductive qualitative coding. Across these domains, surprisal curves produce interpretable classification signals with clear minima near expected ordinal scale positions, and entropy over the completion tended to distinguish genuinely ambiguous items from easier items.