Reproduce “Receptiveness, Not Sycophancy: Distinguishing Engagement from Deference in Language Models” 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

Calvin Isley, Johann Gaebler, Max Lamparth, Julia Minson, Sharad Goel

The authors examine whether evaluations of social sycophancy can mistake receptive disagreement for deference in Reddit moral-advice responses. They compare human and model replies, receptive rewrites, participant ratings, and model prompting strategies that preserve judgments. Participants preferred receptive rewrites, while anchored revision increased receptiveness with little deference change in two models; findings concern tested moral-advice settings.

Summary by alphaXiv.

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