BriefGPT.xyz
Jun, 2024
情感回应生成的理论驱动数据集构建与优化
EmPO: Theory-Driven Dataset Construction for Empathetic Response Generation through Preference Optimization
HTML
PDF
Ondrej Sotolar
TL;DR
我们提出了一种新方法，利用理论驱动的偏好数据集和优化算法来对齐LLMs，以提高情感回应生成的质量和模型的泛化性能，并通过EmpatheticDialogues数据集以及diff-EPITOME和BERTscore指标来评估其效果。
Abstract
empathetic response generation
is a desirable aspect of
conversational agents
, crucial for facilitating engaging and emotionally intelligent multi-turn conversations between humans and machines. Leveraging
→