How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative Reasoning

TL;DR

This paper investigates how large language models read and interpret reasoning traces during quantitative reasoning, revealing patterns linked to correctness and proposing a training-free steering method to improve accuracy.

Contribution

It uncovers self-reading patterns associated with correct reasoning and introduces a novel, training-free steering approach based on Self-Reading Quality scores to enhance model performance.

Findings

Benign self-reading patterns correlate with correct answers.

Incorrect solutions show diffuse and irregular attention.

The proposed SRQ-based steering improves inference accuracy.

Abstract

Thinking LLMs produce reasoning traces before answering. Prior activation steering work mainly targets on shaping these traces. It remains less understood how answer tokens actually read and integrate the reasoning to produce reliable outcomes. Focusing on quantitative reasoning, we analyze the answer-to-reasoning attention and observe a benign self-reading pattern aligned with correctness, characterized by a forward drift of the reading focus along the reasoning trace and a persistent concentration on key semantic anchors, whereas incorrect solutions exhibit diffuse and irregular attention pattern. We interpret this as internal certainty during answer decoding, where the model commits to a viable solution branch and integrates key evidence. Following this, we propose a training-free steering method driven by Self-Reading Quality (SRQ) scores combining geometric metrics for process control with semantic metrics for content monitoring. SRQ selects data to build steering vectors that guide inference toward benign self-reading and away from uncertain and disorganized reading. Experiments show that our method yields consistent accuracy gains.