Reproduce “Not All Information Is Worth Asking For: Decision-Informed Information Seeking with LLMs” 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

Seunghan Lee, Jun Seo, Junhyeok Kang, Jaehoon Lee, Sangjun Han, Sungdong Yoo, Minjae Kim, Tae Yoon Lim and 4 more

DIVA evaluates whether information checks are worth their cost and when to stop. It elicits numeric estimates of risk and check reliability, then uses Bayes’ rule to assess how results could change the best action. On SEC Ret-1d, DIVA-U averages 0.67 first-decision regret versus 1.26 for Direct prompting across 20 LLMs; scores use historical reference models, not realized investment outcomes.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.