Events

Towards Accountable Conversational Agents for Task Completion

11:40 AM to 12:40 PM

CSB 451 CS Auditorium

Dilek Hakkani-Tür, University of Illinois Urbana-Champaign

Abstract:
Task-oriented dialogue systems are designed to assist users in achieving specific, well-defined goals or tasks through natural language interactions. These systems act as a conversational bridge, connecting users to task-specific APIs or tools. Recent advancements have resulted in a significant paradigm shift with the integration of Large Language Models (LLMs) augmented with tool-calling into task-oriented dialogue systems and user simulators that are used for model evaluation and training. However, several issues remain that ponder the use of these systems in real applications.

In this talk, I will share our latest research towards accountability in multi-turn interactions with Large Language Models (LLMs). Our approach, called Reasoning, Acting, and Speaking (ReSpAct), has demonstrated higher task completion rates by engaging with users. To counter user over-reliance, we developed specialized accountability models for dialogue state tracking errors in task-oriented dialogue systems. We also adapted existing annotated dialogue datasets to train an LLM proficient in both interaction and tool calling, showcasing its performance on dialogue system and agentic LLM benchmarks. Our subsequent work utilized reinforcement learning for tool-based reasoning, introducing novel reward mechanisms. Finally, I will discuss user simulation for model evaluation and training. We observed that LLM-based user simulators can deviate from user goals over multi-turn interactions. To address this, we proposed a novel framework that tracks user goal progression throughout conversations, enabling the creation of user simulators that can autonomously monitor goal progression and reason to generate goal-aligned responses.