Reproduce “When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows” 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
Yuchen Ling, Shengcheng Yu, Zhenyu Chen, Chunrong Fang
Jev replaces an agent’s action-selection model with a service that selects among agent-supplied candidates, while the workflow handles observations, execution, and auxiliary calls. Across five agent families in text-only household and mobile benchmarks, it reduces action-selection cost 66.1%–78.4% and time 73.9%–81.5%. Task outcomes vary: T3A’s AndroidWorld success falls from 57.47% to 44.25%; findings are limited to tested integrations and tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.