Reproduce “EvoX: Rethinking the Agent Harness as a Runtime Policy” 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
Haoyang Zhang, Qumeng Sun, Xiao Zhang, Xiang Liu, Zhidong Fan, Junjie Wang
EvoX separates model reasoning from runtime control to help execute tool calls, recover from errors, and ground completion reports. Its Action Realization, Execution Continuity, and Outcome Grounding mechanisms operate as a runtime design. In the non-interactive baseline, EvoX achieved 56.2% task success across 691 task–model pairs, versus 53.1% for the other harnesses’ mean; the lead is not statistically significant or universal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.