Reproduce “Improving Proactive AI Assistance with Hierarchical Procedural Understanding” 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
Jin-Seop Lee, TaeYeon Won, SeongJun Jung, JungHoon Kim, Boyang Albert Li, JinYeong Bak, Jaehong Yoon, Jee-Hyong Lee
ProactiveCoach trains linked phase, step, and action guidance: fine events track progress, broad goals provide context. A lightweight router displays the requested level. On ProactiveCoachBench’s step-level task, Qwen3.5-4B averages 59.30 with three-level supervision versus 50.06 with step-only training. Across four transitions, routing beats in-context switching by 57.1 points, using predefined anchors and ground-truth dialogue history.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.