Reproduce “OpenGuanDan: A Large-Scale Imperfect Information Game Benchmark” 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
Chao Li, Shangdong Yang, Chiheng Zhan, Zhenxing Ge, Yujing Hu, Bingkun Bao, Xingguo Chen, Yang Gao
OpenGuanDan provides a benchmark for GuanDan, a four-player cooperative-competitive card game, by packaging a simulator and independent per-player interfaces that expose observations and legal actions. It compares existing rule-based and learning-based agents in pairwise games and against volunteers. Learning-based agents outperform the included rule-based agents, but GS2 wins 42.5% of 106 volunteer rounds; no professionals participated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.