TriEx: A Game-based Tri-View Framework for Explaining Internal Reasoning in Multi-Agent LLMs
This work addresses explainability for LLM agents in multi-agent environments, offering a novel framework for evidence-grounded evaluation, though it is incremental in applying structured methods to a known bottleneck.
The paper tackles the challenge of explaining Large Language Model (LLM) agents in interactive, partially observable settings by introducing TriEx, a tri-view framework that structures explanations into aligned artifacts, enabling scalable analysis of faithfulness, belief dynamics, and evaluator reliability.
Explainability for Large Language Model (LLM) agents is especially challenging in interactive, partially observable settings, where decisions depend on evolving beliefs and other agents. We present \textbf{TriEx}, a tri-view explainability framework that instruments sequential decision making with aligned artifacts: (i) structured first-person self-reasoning bound to an action, (ii) explicit second-person belief states about opponents updated over time, and (iii) third-person oracle audits grounded in environment-derived reference signals. This design turns explanations from free-form narratives into evidence-anchored objects that can be compared and checked across time and perspectives. Using imperfect-information strategic games as a controlled testbed, we show that TriEx enables scalable analysis of explanation faithfulness, belief dynamics, and evaluator reliability, revealing systematic mismatches between what agents say, what they believe, and what they do. Our results highlight explainability as an interaction-dependent property and motivate multi-view, evidence-grounded evaluation for LLM agents. Code is available at https://github.com/Einsam1819/TriEx.