1
Clarify, Abstain or Answer? Strategising in Conversation with Belief-Augmented Generation
论文提出一种信念增强的对话生成策略，让模型在对话中主动选择澄清、拒绝或直接回答，提升智能交互的可靠性。
arXiv:2605.25831v1 Announce Type: cross Abstract: Large language models (LLMs) define a distribution over text, which can be viewed as a probabilistic…