Reproduce “Recursive Environment Scaling for General Long-Horizon 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

Lingxiao Jing, Zhengbo Jiao, Yichen Xiang, Bofan Zhu, Lingbei Meng, Yan Ni, Haorui Chen, Ran Ran and 5 more

Recursive Environment Scaling (RES) makes existing executable benchmarks more demanding by changing information access and dependencies while preserving task goals and completion requirements. It validates transformed environments through replayed operations and evaluates agents in paired attempts. Across six models and three benchmarks, success declined in all 18 comparisons; DeepSeek-V4.1-Flash fell from 82.86% to 5.71% on Claw-Eval-Live.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.