Reproduce “Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition” 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
Wenjie Lou, Tianshuo Peng, Shiyang Feng, Shengji Tang, Runmin Ma, Ruizhe Chen, Jinxin Shi, Haoyang Peng and 91 more
Sage-A1.5 trains an agent to organize long-horizon research by delegating parallel investigations and reviewing evidence, assumptions, and stalled searches before changing its plan. In a same-teacher BrowseComp comparison, verification mode scored 85.5 accuracy versus 81.0 in normal mode; paired attempts also used fewer output tokens and turns. The training remains centered on search and research synthesis, not validated physical experiments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.