Evaluating and Enhancing Out-of-Domain Generalization of Task-Oriented Dialog Systems for Task Completion without Turn-level Dialog Annotations

2025-02-18

AI Summary

This study addresses the challenge of improving task completion in unseen domains for task-oriented dialogue systems without per-turn state or policy annotations. We propose ZeroToD, a framework integrating schema-aware prompt augmentation, zero-shot prompt contrastive learning, and lightweight supervised fine-tuning—trained solely on natural-language dialogue data, without API call traces or slot annotations. Its core innovation is a schema-enhanced prompting mechanism that significantly improves API call accuracy and cross-domain generalization under zero-label conditions. Evaluated on open-source LLMs—including Llama, Phi, and Qwen—ZeroToD boosts task success rate on unseen domains from a baseline of 53% to substantially higher levels; even small fine-tuned models outperform large proprietary LLMs. Human evaluation confirms that generated responses are informative, fluent, and task-effective.

Technology Category

Machine Learning: Large Multimodal Models (LMMs)Natural Language Processing: Prompt Engineering / PromptingPlanning, Routing, and Scheduling: Planning with Language Models

Application Category

Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved informationEconomics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labelingSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Traditional task-oriented dialog (ToD) systems rely heavily on labor-intensive turn-level annotations, such as dialogue states and policy labels, for training. This work explores whether large language models (LLMs) can be fine-tuned solely on natural language dialogs to perform ToD tasks, without requiring such annotations. We evaluate their ability to generalize to unseen domains and compare their performance with models trained on fully annotated data. Through extensive experiments with three open-source LLMs of varying sizes and two diverse ToD datasets, we find that models fine-tuned without turn-level annotations generate coherent and contextually appropriate responses. However, their task completion performance - measured by accurate execution of API calls - remains suboptimal, with the best models achieving only around 53% success in unseen domains. To improve task completion, we propose ZeroToD, a framework that incorporates a schema augmentation mechanism to enhance API call accuracy and overall task completion rates, particularly in out-of-domain settings. We also compare ZeroToD with fine-tuning-free alternatives, such as prompting off-the-shelf LLMs, and find that our framework enables smaller, fine-tuned models that outperform large-scale proprietary LLMs in task completion. Additionally, a human study evaluating informativeness, fluency, and task completion confirms our empirical findings. These findings suggest the feasibility of developing cost-effective, scalable, and zero-shot generalizable ToD systems for real-world applications.

Problem

• Generalization of task-oriented dialog systems
• Reducing dependency on turn-level annotations
• Enhancing task completion accuracy in unseen domains

Innovation

• LLMs fine-tuned on natural dialogs
• ZeroToD enhances API call accuracy
• Outperforms large-scale proprietary LLMs