Paper ID: 2407.05674

LLM-Based Open-Domain Integrated Task and Knowledge Assistants with Programmable Policies

Harshit Joshi, Shicheng Liu, James Chen, Robert Weigle, Monica S. Lam

Programming LLM-based knowledge and task assistants that faithfully conform to developer-provided policies is challenging. These agents must retrieve and provide consistent, accurate, and relevant information to address user's queries and needs. Yet such agents generate unfounded responses ("hallucinate"). Traditional dialogue trees can only handle a limited number of conversation flows, making them inherently brittle. To this end, we present KITA - a programmable framework for creating task-oriented conversational agents that are designed to handle complex user interactions. Unlike LLMs, KITA provides reliable grounded responses, with controllable agent policies through its expressive specification, KITA Worksheet. In contrast to dialog trees, it is resilient to diverse user queries, helpful with knowledge sources, and offers ease of programming policies through its declarative paradigm. Through a real-user study involving 62 participants, we show that KITA beats the GPT-4 with function calling baseline by 26.1, 22.5, and 52.4 points on execution accuracy, dialogue act accuracy, and goal completion rate, respectively. We also release 22 real-user conversations with KITA manually corrected to ensure accuracy.

Submitted: Jul 8, 2024