Reproduce “HALO: Human-Grounded Granular Evaluation of Proactive 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
Wenqi Pei, Xinyu Wang, Jinrui Liu, Cynthia Zhuge, Jie Sheng, Zihan Wang, Tenghao Huang, Muhao Chen
The authors introduce HALO, a benchmark for testing when personal agents should help. Agents inspect events and retrieve local records, then choose among observing, notifying, preparing, executing, or doing nothing; human annotators judge acceptability and diagnose failures. In its constructed scenarios, 2,188 of 3,575 rejected responses were premature or excessive, a finding not estimating real-world prevalence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.