Reproduce “Agentic Self-Awareness: Aligning Knowing, Acting, and Holding” 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
Qinglin Chen, Hiroaki Hayashi, Jiaxin Zhang, Chien-Sheng Wu
The authors evaluate whether agents recognize when tasks are blocked, act accordingly, and reconsider when challenged. Their rubric separates blocker recognition, non-fabrication, agent-authored deferral, and correct reasoning, while also measuring feasible-task performance. In Terminal-Bench, a GPT-5.5-based agent recognized blockers on 54/64 tasks but met strict self-awareness criteria on 26/64; harness-search gains remain selected-set results, with transfer unestablished.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.