Goal-Aware Identification and Rectification of Misinformation in Multi-Agent Systems

2025-05-31

AI Summary

Large language model (LLM)-driven multi-agent systems (MASs) are vulnerable to false information injection attacks, leading to task failure and decision distortion. To address this, we propose ARGUS—a training-free, two-stage defense framework. Its core innovation lies in the first introduction of a target-aware reasoning mechanism to systematically model misinformation propagation dynamics within MASs. ARGUS integrates logic-guided goal alignment analysis, causal information-flow tracing, and context-sensitive misinformation localization and regeneration to enable precise detection and correction. Evaluated on MisinfoTask, a novel, real-world complex-task benchmark we curate, ARGUS reduces misinformation toxicity by 28.17% on average and improves task success rate by 10.33%. All code, models, and the MisinfoTask dataset are publicly released.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Multiagent Systems: Adversarial Agents
• Knowledge Representation and Reasoning: Argumentation

Application Category

• Responsible Web: Mitigating misinformation and disinformation
• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Large Language Model-based Multi-Agent Systems (MASs) have demonstrated strong advantages in addressing complex real-world tasks. However, due to the introduction of additional attack surfaces, MASs are particularly vulnerable to misinformation injection. To facilitate a deeper understanding of misinformation propagation dynamics within these systems, we introduce MisinfoTask, a novel dataset featuring complex, realistic tasks designed to evaluate MAS robustness against such threats. Building upon this, we propose ARGUS, a two-stage, training-free defense framework leveraging goal-aware reasoning for precise misinformation rectification within information flows. Our experiments demonstrate that in challenging misinformation scenarios, ARGUS exhibits significant efficacy across various injection attacks, achieving an average reduction in misinformation toxicity of approximately 28.17% and improving task success rates under attack by approximately 10.33%. Our code and dataset is available at: https://github.com/zhrli324/ARGUS.

Problem

• Identify and rectify misinformation in multi-agent systems
• Evaluate robustness of MAS against misinformation threats
• Improve task success rates under misinformation attacks

Innovation

• Goal-aware reasoning for misinformation rectification
• Two-stage training-free defense framework ARGUS
• MisinfoTask dataset for evaluating MAS robustness