CORE: Measuring Multi-Agent LLM Interaction Quality under Game-Theoretic Pressures
This provides a diagnostic tool for measuring linguistic robustness in multi-agent LLM systems, addressing a gap in quantifying interaction quality, though it is incremental as it builds on existing game-theoretic and linguistic analysis methods.
The paper tackles the problem of quantifying linguistic diversity in multi-agent LLM interactions under game-theoretic pressures by introducing CORE, a metric that integrates cluster entropy, lexical repetition, and semantic similarity, and finds that cooperative settings show more repetition and vocabulary expansion while competitive ones have less repetition and constrained vocabularies.
Game-theoretic interactions between agents with Large Language Models (LLMs) have revealed many emergent capabilities, yet the linguistic diversity of these interactions has not been sufficiently quantified. In this paper, we present the Conversational Robustness Evaluation Score: CORE, a metric to quantify the effectiveness of language use within multi-agent systems across different game-theoretic interactions. CORE integrates measures of cluster entropy, lexical repetition, and semantic similarity, providing a direct lens of dialog quality. We apply CORE to pairwise LLM dialogs across competitive, cooperative, and neutral settings, further grounding our analysis in Zipf's and Heaps' Laws to characterize word frequency distributions and vocabulary growth. Our findings show that cooperative settings exhibit both steeper Zipf distributions and higher Heap exponents, indicating more repetition alongside greater vocabulary expansion. In contrast, competitive interactions display lower Zipf and Heaps exponents, reflecting less repetition and more constrained vocabularies. These results provide new insights into how social incentives influence language adaptation, and highlight CORE as a robust diagnostic for measuring linguistic robustness in multi-agent LLM systems. Our code is available at https://github.com/psyonp/core.