Reproduce “Object-Oriented Harness: Evolving Agent Harnesses through Object-Oriented Design” 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
Gaotang Li, Bhavana Dalvi Mishra, Yifei Ming, Yanfei Chen, Zhongying CuiZhu, Lesly Miculicich, Long T. Le, Rui Meng and 4 more
OO-Harness evolves specialized agent harnesses as subclasses, inheriting or overriding behavior, then routes tasks using task information and observed Base execution. On AutomationBench, its held-out mean partial credit was 77.41%, versus 75.40% for Adaptive Auto-Harness and 61.62% for Base. It had the highest primary score on all three benchmarks. In the application-disjoint test, the router selected Base for all 49 tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.