AEMA: Verifiable Evaluation Framework for Trustworthy and Controlled Agentic LLM Systems

AAAI 2026 Workshop TrustAgent Submission16 Authors

Keywords: Agentic AI, Multi-Agent Systems, Trustworthy AI, Verifiable Evaluation, Human Oversight

TL;DR: AEMA is a framework that makes multi-agent LLM systems trustworthy and auditable by evaluating their decisions through verifiable, reproducible, and human-aligned processes.

Abstract: Evaluating large language model (LLM)-based multi-agent systems remains a critical challenge, as these systems must exhibit reliable coordination, transparent decision-making, and verifiable performance across evolving tasks. Existing evaluation approaches often limit themselves to single-response scoring or narrow benchmarks, which lack stability, extensibility, and automation when deployed in enterprise settings at multi-agent scale. We present \textbf{AEMA (Adaptive Evaluation Multi-Agent)}, a process-aware and auditable framework that plans, executes, and aggregates multi-step evaluations across heterogeneous agentic workflows under human oversight. Compared to a single LLM-as-a-Judge, AEMA achieves greater stability, human alignment, and traceable records that support accountable automation. Our results on enterprise-style agent workflows simulated using realistic business scenarios demonstrate that AEMA provides a transparent and reproducible pathway toward responsible evaluation of LLM-based multi-agent systems.

Submission Number: 16