Single-Turn Evaluation Paradigm

An evaluation methodology for AI dialogue systems that assesses a model's response quality given a single user input. It offers strong reproducibility and low cost but cannot capture cross-turn sequential failures, such as commitment fulfillment rates.