What question did this study set out to answer?

This research aims to improve empathetic dialogue generation by leveraging large language model preference signals.

June 11, 2026

Pairwise Empathetic Dialogue Generation via Automatic Large Language Model Preference Data Construction

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial demonstrates improved empathetic dialogue generation in automated agents, suggesting enhanced emotional interactions.

This research aims to improve empathetic dialogue generation by leveraging large language model preference signals.
Developed a pairwise learning framework utilizing a large language model to generate multiple candidate responses.
Refined candidates using Maximal Marginal Relevance and Jaccard similarity scoring.
Evaluated responses based on empathy, fluency, relevance, and reliability using a secondary language model to extract preference pairs.
The proposed model achieved superior performance in fluency (PPL) and diversity (Distinct-1/2) metrics compared to existing models.
Empathetic alignment was enhanced, with an improved Emp-F1 score confirmed through automatic preference data generation.

Lim et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark