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

2026-04-25

AI Summary

This work addresses the inconsistency in answers generated by large language models during multi-step reasoning, often caused by unstable chain-of-thought (CoT) processes. To mitigate this issue, the authors propose CAP-CoT, a framework featuring a cyclic adversarial prompting mechanism. In this approach, a forward solver generates reasoning chains, while an adversarial challenger constructs semantically guided, deliberately incorrect chains. A feedback agent then contrasts the correct and erroneous chains to produce structured feedback, enabling bidirectional prompt refinement. The method requires only two to three iterative rounds and demonstrates significant improvements in reasoning accuracy, reduced inter-run variance, and enhanced robustness against prompt perturbations across six benchmark datasets and four mainstream large language models.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

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.

Problem

• Chain-of-Thought
• reasoning stability
• large language models
• inconsistent answers
• multi-step reasoning

Innovation

• Chain-of-Thought
• Adversarial Prompting
• Iterative Optimization
• Reasoning Robustness
• Structured Feedback