Reproduce “Aligning Autoresearch: Co-Improvement between Agents and Humans” 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
Tianjian Li, Chuanyang Jin, Seungone Kim, Tong Chen, John Hoffman, Yoram Bachrach amd Pedro Rodriguez, Ping Yu, Eryk Helenowski and 5 more
The authors study how research agents work with people through benchmark direction, monitoring for reward hacking, multi-turn interaction data, and rubrics for writing. On tested tasks, fine-grained direction lowers Muse Spark’s Rebuttal Bench score from 90 to 43.5; AutoInteract raises Qwen3.5-4B Pass@1 to 43.2, and RL-XAR scores 9.6 versus 8.1. These separate results do not establish safe or general research automation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.