Measuring AI Reasoning: A Guide for Researchers

TL;DR

This paper advocates for evaluating AI reasoning through intermediate, process-oriented measures like reasoning traces, rather than solely relying on final-answer accuracy, to better diagnose and improve language models.

Contribution

It formalizes reasoning as a search-like process, highlights limitations of current models, and proposes process-based evaluation using reasoning traces as a new standard.

Findings

Final-answer accuracy is insufficient for diagnosing reasoning processes.

Intermediate decoding and external reasoning traces improve evaluation.

Scalable architectures are limited in variable-depth reasoning capabilities.

Abstract

In this paper, we offer a guide for researchers on evaluating reasoning in language models, building the case that reasoning should be assessed through evidence of adaptive, multi-step search rather than final-answer accuracy alone. Under an evaluation-oriented definition, reasoning requires selecting intermediate steps and halting according to input-dependent conditions, which we formalize as a search-like procedure. We show that single forward passes in scalable architectures are structurally limited in their ability to realize such variable-depth computation, motivating intermediate decoding and externalized reasoning traces as appropriate evaluation interfaces. Central to our argument is that final-answer accuracy alone is an insufficient measure of reasoning, because it provides little ability to diagnose or debug the underlying processes that produce individual solutions in frontier models. We therefore argue for a shift toward process-based evaluation, in which reasoning is assessed through the faithfulness and validity of intermediate reasoning traces as first-class evaluation targets.