Collective Bias Mitigation via Model Routing and Collaboration

2026-10-02

AI Summary

This study addresses the inherent difficulty of mitigating deep-seated biases in large language models (LLMs) due to their reliance on singular knowledge sources. To overcome this limitation, this work proposes a collective bias mitigation framework that systematically explores the effective selection and organization of multiple LLMs. Grounded in fine-grained model behavior learning, the framework facilitates multi-agent collaboration and knowledge sharing through a model routing mechanism integrated with communication topologies such as debate and committee structures. The proposed approach achieves significantly fairer response generation compared to independent baseline models, notably reducing age bias scores from 0.25 to 0.10. Furthermore, it effectively balances debiasing performance with inference costs, demonstrating the potential of organized multi-LLM systems for robust bias mitigation.

Abstract

Large language models (LLMs) are increasingly deployed in public health, finance, and governance, requiring both accuracy and societal value alignment. Despite recent advances, LLMs often perpetuate or amplify bias embedded in their training data, posing challenges to fairness. While self-debiasing encourages an LLM to identify and correct its own biases, relying on a single model's intrinsic knowledge may be insufficient to address deeply ingrained stereotypes. To address this limitation, we introduce Collective Bias Mitigation (CBM), a framework that alleviates bias by learning fine-grained model behavior and fostering knowledge sharing among diverse LLMs. This work is the first to systematically explore the effective selection and organization of distinct LLMs to cultivate fairer LLM responses. Experiments show CBM substantially outperforms standalone baselines (e.g., in the top-7 setting, Committee lowers the age bias score from 0.25 to 0.10). Our Debating and Committee topologies achieve substantial bias reduction, with the latter balancing mitigation effectiveness and inference cost, highlighting the potential of CBM for fairer LLMs.

Problem

• Large Language Models
• Bias Mitigation
• Fairness
• Collective Intelligence
• Stereotypes

Innovation

• Collective Bias Mitigation
• Model Routing
• Large Language Models
• Multi-Model Collaboration
• Fairness