CoE: Collaborative Entropy for Uncertainty Quantification in Agentic Multi-LLM Systems

📰 ArXiv cs.AI

Collaborative Entropy (CoE) is a metric for uncertainty quantification in multi-LLM systems, addressing semantic disagreement across models

advanced Published 31 Mar 2026

Action Steps

Define a shared semantic cluster space across multiple LLMs
Calculate intra-model semantic uncertainty
Calculate inter-model semantic disagreement
Combine intra-model and inter-model uncertainties using CoE metric

Who Needs to Know This

AI engineers and researchers working on multi-LLM systems can benefit from CoE to better understand and quantify uncertainty in their models, enabling more accurate and reliable predictions

Key Insight

💡 CoE captures semantic disagreement across models, providing a more comprehensive understanding of uncertainty in multi-LLM systems

Key Takeaways

Collaborative Entropy (CoE) is a metric for uncertainty quantification in multi-LLM systems, addressing semantic disagreement across models

Full Article

Title: CoE: Collaborative Entropy for Uncertainty Quantification in Agentic Multi-LLM Systems

Abstract:
arXiv:2603.28360v1 Announce Type: new Abstract: Uncertainty estimation in multi-LLM systems remains largely single-model-centric: existing methods quantify uncertainty within each model but do not adequately capture semantic disagreement across models. To address this gap, we propose Collaborative Entropy (CoE), a unified information-theoretic metric for semantic uncertainty in multi-LLM collaboration. CoE is defined on a shared semantic cluster space and combines two components: intra-model sem

Read full paper → ← Back to Reads