Reproduce “Reflecting Twice before Speaking with Empathy: Self-Reflective Alternating Inference for Empathy-Aware End-to-End Spoken Dialogue” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuhang Jia, Pei Liu, Haoqin Sun, Jiaming Zhou, Xuxin Cheng, Cao Liu, Ke Zeng, Xunliang Cai and 1 more
ReEmpathy trains a spoken dialogue model to interleave unspoken empathy-related reflection with response speech chunks, using descriptive assessments to guide fine-tuning. On the Chinese OpenS2S evaluation, its EmpathyEval average—the mean of four 1–5 ratings—was 2.9633, versus 2.9393 without reflection and 2.9555 for supervised fine-tuning. These differences do not establish user preference or reflection as the sole cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.