DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses

Key points

• Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging.1
• Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning.1
• We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs.1
• DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths.1

Sources (2)

• [1]DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses
  Hugging Face Daily Papers · Oct 16, 12:00 AM

  We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations.

  Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging.

• [2]DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 05:56 PM · same content

Extractive summary: sentences quoted from the sources.