Game Theory Driven Multi-Agent Framework Mitigates Language Model Hallucination

2026-07-09

AI Summary

This work addresses the susceptibility of lightweight large language models to hallucination and their limited capacity for axiomatic reasoning in rigorously structured scientific domains. To overcome these challenges, the authors propose G-Frame, a multi-agent framework that, for the first time, integrates game-theoretic adaptive teaming with Bayesian inference to internalize domain-specific constraints through structured collaboration. This approach enables high-quality chain-of-thought synthesis and establishes a closed-loop training pipeline, generating 360,000 reasoning chains and nearly 200,000 question-answer pairs. The resulting OmniChem-7B model achieves performance on par with GPT-4o mini on benchmarks such as ChemBench, while reducing hallucination rates by 79.46%. Moreover, it demonstrates notable capabilities in molecular design and synthesis planning, offering a scalable new paradigm for scientific knowledge discovery.

Abstract

The application of lightweight Large Language Models in rule-based scientific domains remains severely limited by their tendency to mimic linguistic patterns rather than reproduce axiomatic reasoning, causing frequent hallucinations. Here, we show that G-Frame, an adaptive multi-agent framework integrating Bayesian and team game principles, establishes an automated closed-loop for high-quality data synthesis and model training. By forcing the internalization of domain constraints through structured reasoning, we synthesized a specialized corpus of 363,045 chains-of-thought and 199,589 question-answer pairs. The resulting 7B model OmniChem achieves performance parity with GPT 4o mini on custom benchmarks and ChemBench while exhibiting a 79.46% reduction in hallucinations relative to its base architecture. We further demonstrate the advanced capabilities of OmniChem in molecular design and synthesis planning. This work establishes a scalable paradigm utilizing adaptive multi-agents to overcome inherent reasoning deficiencies, offering a feasible pathway for accelerating knowledge discovery in specialized scientific fields.

Problem

• hallucination
• large language models
• scientific reasoning
• axiomatic reasoning
• domain constraints

Innovation

• multi-agent framework
• game theory
• hallucination mitigation
• structured reasoning
• Bayesian inference