Reproduce “Emergence of Reputation-Based Cooperation in LLM Agents” 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
Kazuya Horibe, Kenji Itao, Wataru Toyokawa
The authors study whether evolved donation strategies in LLM agents resist exploitation by an always-zero-donation bot. Twelve agents revise natural-language strategies from observed behavior across 10 generations, then 11 copies of each strategy face the bot in a one-generation invasion test. Opponent endowment sensitivity explained 93.4% of robustness variation across four models, but this result is model-level and test-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.