CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning

TL;DR

CAP-CoT introduces a cycle adversarial prompt framework that enhances reasoning accuracy and stability in large language models by iteratively identifying and correcting logical vulnerabilities in chain-of-thought reasoning.

Contribution

It proposes a novel cycle adversarial prompt method that improves both accuracy and consistency of LLM reasoning through iterative, task-semantic adversarial feedback.

Findings

Reduces variability in reasoning across multiple runs.

Improves reasoning accuracy on six benchmark datasets.

Enhances robustness to prompt perturbations.

Abstract

Chain-of-Thought (CoT) prompting has emerged as a simple and effective way to elicit step-by-step solutions from large language models (LLMs). However, CoT reasoning can be unstable across runs on long, multi-step problems, leading to inconsistent answers for unchanged task. Most prior work focuses on improving the forward reasoning chain within a single pass, with less attention to iterative and contrastive correction. To address this gap, we propose CAP-CoT, a Cycle Adversarial Prompt optimization framework designed to improve both CoT reasoning accuracy and stability of a single deployed solver. In each cycle, a forward solver generates candidate reasoning chains, an adversarial challenger constructs plausible but deliberately flawed chains using targeted error strategies, and a feedback agent contrasts the two chains and produces step-aligned structured feedback. This feedback closes the optimization loop in two directions, including updating the solver prompt based on errors exposed by the challenger, and updating the challenger prompt to generate increasingly targeted errors in subsequent cycles. Unlike safety-oriented adversarial prompting such as jailbreak or prompt-injection attacks, our adversarial component is task-semantic and aims to expose logical vulnerabilities in reasoning chains. Experiments across six benchmarks and four LLM backbones demonstrate that within two to three adversarial prompt optimization cycles, CAP-CoT consistently reduces variability across runs while improving reasoning accuracy and robustness to prompt perturbations.