ATOD: An Evaluation Framework and Benchmark for Agentic Task-Oriented Dialogue Systems
The paper introduces ATOD, a synthetic dataset capturing advanced agentic behaviors like multigoal coordination and proactivity in task-oriented dialogue systems. It also proposes ATOD-Eval, a holistic evaluation framework that systematically assesses these complex agentic capabilities using five key dimensions and an agentic memory system. This provides a unified tool for benchmarking next-generation conversational agents.