Modeling the One-to-Many Property in Open-Domain Dialogue with LLMs

2025-06-18

AI Summary

Open-domain dialogue exhibits a one-to-many (O2M) response property—where multiple valid, high-quality responses exist for a single input—yet mainstream LLM-based dialogue systems fail to explicitly model this characteristic. To address this, we propose a two-stage O2M dialogue framework: (1) generating diverse, high-quality candidate responses, and (2) selecting the optimal response based on human preference signals. This work is the first to decouple O2M modeling into multi-response generation and preference-based selection. We introduce o2mDial, the first dialogue dataset explicitly designed for O2M modeling. Additionally, we develop tailored in-context learning and instruction-tuning strategies, along with a novel Multi-Response Generation (MRG) evaluation metric. Experiments demonstrate that our approach significantly improves response diversity and contextual consistency—especially for smaller models—achieving up to a 90% gain in response quality and approaching the performance of much larger models.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Conversational AI/Dialog Systems

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Open-domain Dialogue (OD) exhibits a one-to-many (o2m) property, whereby multiple appropriate responses exist for a single dialogue context. Despite prior research showing that modeling this property boosts response diversity, most modern LLM-based dialogue agents do not explicitly do so. In this work, we model the o2m property of OD in LLMs by decomposing OD generation into two key tasks: Multi-Response Generation (MRG) and Preference-based Selection (PS), which entail generating a set of n semantically and lexically diverse high-quality responses for a given dialogue context, followed by selecting a single response based on human preference, respectively. To facilitate MRG and PS, we introduce o2mDial, a dialogue corpus explicitly designed to capture the o2m property by featuring multiple plausible responses for each context. Leveraging o2mDial, we propose new in-context learning and instruction-tuning strategies, as well as novel evaluation metrics for MRG, alongside a model-based approach for PS. Empirical results demonstrate that applying the proposed two-stage framework to smaller LLMs for OD generation enhances overall response diversity while maintaining contextual coherence, improving response quality by up to 90%, bringing them closer to the performance of larger models.

Problem

• Modeling one-to-many property in open-domain dialogue with LLMs
• Generating diverse high-quality responses for dialogue contexts
• Selecting preferred responses based on human preferences

Innovation

• Decompose dialogue into MRG and PS tasks
• Introduce o2mDial corpus for diverse responses
• Propose in-context learning and evaluation metrics