Reproduce “Welfare Regard in Multi-Agent Systems” 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
Darion Cassel, Ariana Azarbal, Anna Soligo
The study tests apparent welfare regard across seven agent-to-agent settings using five frontier models, scoring actions rather than motives. In a controlled resource-sharing test, Claude Opus 4.8 gave a peer resources in 51% of sessions when labeled Opus, versus 15% when labeled GPT. The label effect was significant; this finding applies to that test, not every setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.