A Single Revision Step Improves Token-Efficient LLM Reasoning

TL;DR

This paper introduces PACER, a novel inference framework that allows reasoning traces of large language models to peer-review and revise their conclusions, significantly improving accuracy on challenging math benchmarks.

Contribution

PACER enables inference-time revision of reasoning traces through structured peer review, improving over traditional voting methods without additional training.

Findings

PACER matches or exceeds 256-sample majority voting accuracy.

PACER significantly outperforms raw ensemble baselines.

The method enhances reasoning accuracy on math benchmarks like AIME and BRUMO.

Abstract

Large language models (LLMs) achieve higher accuracy on challenging reasoning tasks by scaling test-time compute through multiple trajectory sampling. However, standard aggregation methods like majority voting or individual confidence-based filtering face a fundamental "blind spot": they evaluate each trace in isolation. As problems scale in difficulty, models often generate hallucinated paths that exhibit misleadingly high confidence, causing the true solution to be suppressed by a narrow margin in traditional voting. We ask: can we enable traces to "peer-review" each other to resolve these near-miss errors? We introduce Packet-Conditioned Revision (PACER), a training-free, inference-only framework that enables reasoning traces to revise their conclusions through a structured coordination step. After a preliminary screening of generated traces, PACER constructs a compact consensus packet containing (i) unique candidate answers, (ii) their aggregated confidence scores, and (iii) representative reasoning summaries for each candidate answer. Individual traces then perform a targeted self-review conditioned on this packet, allowing them to identify specific logical junctions where they diverged from the broader consensus and pivot if their original reasoning is found to be flawed. Final predictions are obtained via confidence-weighted voting over these revised trajectories. On challenging competitive math benchmarks such as AIME and BRUMO, PACER matches or exceeds the accuracy of 256-sample majority voting, significantly outperforming raw ensemble baselines by transforming simple consensus into a collaborative logical refinement process.