Multi-LLM Collaboration for Medication Recommendation

TL;DR

This paper introduces a Chemistry-inspired multi-LLM collaboration framework to enhance the reliability, stability, and calibration of medication recommendations in clinical decision support, addressing hallucination and inconsistency issues.

Contribution

It applies a novel Chemistry-based interaction modeling to guide LLM ensembles, improving credibility and stability of AI-driven medication suggestions in healthcare.

Findings

Ensembles are more stable and credible.

Interaction-aware collaboration reduces errors.

Preliminary results show promising improvements.

Abstract

As healthcare increasingly turns to AI for scalable and trustworthy clinical decision support, ensuring reliability in model reasoning remains a critical challenge. Individual large language models (LLMs) are susceptible to hallucinations and inconsistency, whereas naive ensembles of models often fail to deliver stable and credible recommendations. Building on our previous work on LLM Chemistry, which quantifies the collaborative compatibility among LLMs, we apply this framework to improve the reliability in medication recommendation from brief clinical vignettes. Our approach leverages multi-LLM collaboration guided by Chemistry-inspired interaction modeling, enabling ensembles that are effective (exploiting complementary strengths), stable (producing consistent quality), and calibrated (minimizing interference and error amplification). We evaluate our Chemistry-based Multi-LLM collaboration strategy on real-world clinical scenarios to investigate whether such interaction-aware ensembles can generate credible, patient-specific medication recommendations. Preliminary results are encouraging, suggesting that LLM Chemistry-guided collaboration may offer a promising path toward reliable and trustworthy AI assistants in clinical practice.