Reliable Chain-of-Thought via Prefix Consistency

TL;DR

The paper introduces prefix consistency, a reliability signal for Chain-of-Thought reasoning in large language models, improving accuracy and efficiency without needing token probabilities.

Contribution

It proposes a new method called prefix consistency that enhances self-consistency by reweighting answers based on trace stability, requiring no token log-probabilities.

Findings

Prefix consistency outperforms existing correctness predictors across multiple models and benchmarks.

Reweighting votes by prefix consistency achieves accuracy with up to 21x fewer tokens.

The method is effective without access to token log-probabilities or self-rating prompts.

Abstract

Large Language Models often improve accuracy on reasoning tasks by sampling multiple Chain-of-Thought (CoT) traces and aggregating them with majority voting (MV), a test-time technique called self-consistency. When we truncate a CoT partway through and regenerate the remainder, we observe that traces with correct answers reproduce their original answer more often than traces with wrong answers. We use this difference as a reliability signal, prefix consistency, that weights each candidate answer by how often it reappears under regeneration. It requires no access to token log-probabilities or self-rating prompts. Across five reasoning models and four math and science benchmarks, prefix consistency is the best correctness predictor in most settings, and reweighting votes by it reaches Standard MV plateau accuracy at up to 21x fewer tokens (median 4.6x). Our code is available at https://github.com/naoto-iwase/prefix-consistency.