One email each morning — yesterday's AI, sortedGet it in your inbox

METAL LAB

Persona-Guided LLM Agents for Task-Oriented Dialogue

arXiv:2608.180852026-08-20

Prior work has shown that large language models (LLMs) can express diverse personality traits in open-ended text generation. However, it remains unclear whether they can do so in a goal-directed dialogue without compromising task completion, and whether adapting to the user's personality improves the interaction quality. We study these questions in task-oriented dialogue (TOD), where a system helps a user accomplish a goal via multi-turn interactio

Authors · Maryam Shoaeinaeini, Brent Harrison, A. B. Siddique

Read on arXiv

Latest papers

All papers →

Latest from METAL LAB