Reproduce “Empathy Is Not the Scapegoat for Declines in Factuality and Reliability” 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

Letian Shi, Kanwal Mehreen, Deheng Zhang, Yuxia Wang

The authors test how emotional context and tone-focused fine-tuning affect language-model reliability across benchmark tasks. They compare ShareGPT fine-tuning with original, warm, and domain-adaptive responses, and add manually written emotional cues to unchanged questions. Emotional context improves MASK honesty in all six configurations, while medical-reasoning accuracy falls in five of six; effects vary by task and model.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.