MetaCrit: A Critical Thinking Framework for Self-Regulated LLM Reasoning

TL;DR

MetaCrit is a multi-agent framework inspired by metacognitive theory that enhances large language models' reasoning by improving truthfulness, logical consistency, and safety through modular regulation agents.

Contribution

It introduces a novel multi-agent system for self-regulated reasoning in LLMs, grounded in metacognitive regulation theory, with modular agents that can be integrated into existing models.

Findings

Significantly improves content truthfulness and logical soundness.

Eliminates toxic outputs in LLM responses.

Effective across multiple benchmarks and model architectures.

Abstract

Large language models (LLMs) fail on over one-third of multi-hop questions with counterfactual premises and remain vulnerable to adversarial prompts that trigger biased or factually incorrect responses, which exposes a fundamental deficit in self-regulated reasoning. We propose \textbf{MetaCrit}, a multi-agent framework grounded in Nelson and Narens' metacognitive regulation theory. MetaCrit decomposes reasoning regulation into four agents: object-level generation, a \emph{monitoring} agent that assesses response validity, a \emph{control} agent that critiques logical soundness, and a meta-level synthesizer that integrates all signals into a final response. Evaluation across eight benchmarks, four model backbones, and a college-level analytical writing study shows that MetaCrit significantly improves content truthfulness and logical soundness while eliminating toxic outputs. Its modular design allows individual agents to be integrated into existing frameworks as drop-in components without architectural modifications.