Reproduce “AutoRefine: Compiling Trajectories into Validated Typed Agent Artifacts” 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
Libin Qiu, Zhirong Gao, Junfu Chen, Yuhang Ye, Liangyu Li, Weizhi Huang, Xiaobo Xue, Wenkai Qiu and 1 more
AutoRefine turns agent experience into Rules for local decisions, Skills for parent-context procedures, or Subagents for bounded objectives needing isolated state and completion conditions. It derives evidence-linked specifications, then applies contract checks and replay against correction and preservation cases. On TravelPlanner, it passes 145 of 180 queries (80.56%), versus AutoManual’s 50.0%; these results do not establish global non-regression.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.