BriefGPT.xyz
Oct, 2021
SaFeRDialogues: 对话安全故障后优雅地接受反馈
SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures
HTML
PDF
Megan Ung, Jing Xu, Y-Lan Boureau
TL;DR
该论文提出了SaFeRDialogues——关于安全问题的对话反馈的优雅响应的任务和数据集，通过在数据集上进行微调，使得基于该方法的模型生成的对话可大大提高语境感知，更具人性化。
Abstract
Current open-domain
conversational models
can easily be made to talk in inadequate ways.
online learning
from conversational
feedback
give
→