1 paper
Maria Emilia Mazzolenis, Ruirui Zhang
Large language models (LLMs) are increasingly applied in task-oriented dialogue (TOD) systems but often struggle with long, conditional workflows that involve external tool calls a…