Finding the Cracks: Improving LLMs Reasoning with Paraphrastic Probing and Consistency Verification

TL;DR

This paper introduces PPCV, a framework that improves large language models' reasoning by identifying critical tokens through paraphrastic probing and verifying answer consistency across multiple reasoning paths.

Contribution

The paper presents a novel two-stage method for detecting and leveraging critical tokens to enhance reasoning accuracy in LLMs, addressing challenges in error accumulation.

Findings

PPCV significantly improves reasoning performance on multiple benchmarks.

The method effectively identifies influential tokens in reasoning processes.

Enhanced reasoning accuracy compared to baseline models.

Abstract

Large language models have demonstrated impressive performance across a variety of reasoning tasks. However, their problem-solving ability often declines on more complex tasks due to hallucinations and the accumulation of errors within these intermediate steps. Recent work has introduced the notion of critical tokens--tokens in the reasoning process that exert significant influence on subsequent steps. Prior studies suggest that replacing critical tokens can refine reasoning trajectories. Nonetheless, reliably identifying and exploiting critical tokens remains challenging. To address this, we propose the Paraphrastic Probing and Consistency Verification~(PPCV) framework. PPCV operates in two stages. In the first stage, we roll out an initial reasoning path from the original question and then concatenate paraphrased versions of the question with this reasoning path. And we identify critical tokens based on mismatches between the predicted top-1 token and the expected token in the reasoning path. A criterion is employed to confirm the final critical token. In the second stage, we substitute critical tokens with candidate alternatives and roll out new reasoning paths for both the original and paraphrased questions. The final answer is determined by checking the consistency of outputs across these parallel reasoning processes. We evaluate PPCV on mainstream LLMs across multiple benchmarks. Extensive experiments demonstrate PPCV substantially enhances the reasoning performance of LLMs compared to baselines.