Metacognitive Self-Correction for Multi-Agent System via Prototype-Guided Next-Execution Reconstruction

2025-10-16

AI Summary

To address the vulnerability of error cascading in large language model–driven multi-agent systems (MAS), this paper proposes an unsupervised metacognitive self-correction framework. The method models anomaly detection as a history-conditioned execution step reconstruction task, innovatively incorporating prototype priors to enhance representation stability under sparse contextual conditions. It further integrates causal consistency modeling with an anomaly scoring mechanism to enable fine-grained, real-time, step-level error identification and targeted correction. A correction agent dynamically intervenes via next-step reconstruction and prototype-guided embedding learning. Evaluated on the Who&When benchmark, the framework achieves an 8.47% improvement in error detection AUC-ROC and delivers consistent end-to-end performance gains across diverse MAS architectures, while incurring negligible computational overhead.

Technology Category

• Multiagent Systems: Adversarial Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Large Language Model based multi-agent systems (MAS) excel at collaborative problem solving but remain brittle to cascading errors: a single faulty step can propagate across agents and disrupt the trajectory. In this paper, we present MASC, a metacognitive framework that endows MAS with real-time, unsupervised, step-level error detection and self-correction. MASC rethinks detection as history-conditioned anomaly scoring via two complementary designs: (1) Next-Execution Reconstruction, which predicts the embedding of the next step from the query and interaction history to capture causal consistency, and (2) Prototype-Guided Enhancement, which learns a prototype prior over normal-step embeddings and uses it to stabilize reconstruction and anomaly scoring under sparse context (e.g., early steps). When an anomaly step is flagged, MASC triggers a correction agent to revise the acting agent's output before information flows downstream. On the Who&When benchmark, MASC consistently outperforms all baselines, improving step-level error detection by up to 8.47% AUC-ROC ; When plugged into diverse MAS frameworks, it delivers consistent end-to-end gains across architectures, confirming that our metacognitive monitoring and targeted correction can mitigate error propagation with minimal overhead.

Problem

• Detecting cascading errors in multi-agent systems
• Enabling real-time unsupervised step-level self-correction
• Mitigating error propagation through prototype-guided reconstruction

Innovation

• Metacognitive framework enables real-time error detection
• Next-Execution Reconstruction predicts causal consistency embeddings
• Prototype-Guided Enhancement stabilizes anomaly scoring process