Reproduce “The Between-Round Decision in Autonomous Research: An LLM Controller That Reads the Whole Trajectory” 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
Caoxin Wang, Shengkai Gan, Xiang Hu, Lili Pan
BRED lets an LLM choose between experimental rounds whether to continue from the best verified result, start a new line, refine, or stop. It bases this choice on the Champion and recorded trajectory. On ETTh1, the best verified MAE was 0.4172; full-history decisions outperformed no-history decisions on two tested tasks, though GENERATE retained full logs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.