Reproduce “RSIHub: A Composable Framework for Agent-Evolution Methods” 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
Zimu Wang, Xiaobo Wang, Siyu Ye, Jiaqi Li, Silin Chen, Yuling Shi, Ruobing Wang, Jingjing Zhang and 6 more
RSIHub provides shared infrastructure for agent evolution, separating reusable policy choices from candidate execution, evaluation, recovery, and lineage. Recipes adapt four methods within a lifecycle, while framework services manage experiments and link outcomes to evaluation conditions. On Terminal Bench 2, MiniSWE reaches 69.7% pass@1, up from 56.2% under Hyperagents and A-Evolve recipes; this result reflects those configurations, not RSIHub alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.