Tag
该论文提出 PeopleMakingConversation 基准，通过 176 段非脚本对话中的 427 个停顿点和 482 个话轮结束点，评估六个实时语音系统在何时开口、延迟、回复长度与对话记忆等方面的自然对话表现，并公开数据与评测框架。
This paper introduces GrowLoop, a self-evolving evaluation system for assessing human-likeness in open-ended conversations. It uses minimal human seed annotations to iteratively refine evaluation rubrics, addressing challenges of tacit knowledge, varying human agreement, and evolving model capabilities.