QUESTIONS WORTH TESTING
Explore the next question.
Trace candidate hypotheses back to evidence, compare model responses, and record expert judgment.
Draft briefs await Raghu’s scientific validation. Candidate ideas are hypotheses, not established novelty. Saving a review does not approve training.
Search scope & limitations
Evidence ledger
Problem briefs
Candidate ideas
Check the closest known work before judging originality or feasibility.
Model experiment
Baseline and adapted models, each with and without supplied evidence. Model identities are visible; these outputs require scientific review.