Reproduce “Environment Duels: A Microcosm of LLM Co-Evolution” 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
Shibo Jin, Zhiqiu Xu, Yiling Zou, Shangyu Gong, Mayur Naik
Environment Duels evaluates models as authors of interactive, executable environments and solvers of other models’ tasks. Authors provide hints and may revise tasks after observing play; scores separately measure solving, difficulty, and net hint lift. Claude-Opus-5 leads solving, while Qwen3.8-27B leads hint lift. For one learner, hint-lift order across four authors matches training gains, without establishing causation or generality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.