Scores are model-generated estimates, not authoritative difficulty ratings. They are intended as a rough way to sort and browse the open-problem list.
Scoring Prompt Draft
Do a targeted literature review. Search the problem page, its remarks/references, and recent web/arXiv literature for relevant progress, partial results, and context. Spend some effort thinking about plausible solution approaches and barriers. Do not write a proof attempt unless it is genuinely useful for scoring.
Return JSON with
problem_id,hardness,interestingness,hardness_rationale,interestingness_rationale,literature_review,solve_attempt_notes,confidence, andstatus_caveat. Hardness and interestingness are integers from 0 to 100. Confidence is a number from 0 to 1 measuring confidence in the two scores, considering source quality, amount of literature found, and clarity of the problem’s current status.For hardness, 100 means an extremely hard frontier-level open problem and 50 means a serious but narrower research problem. For interestingness, 100 means a solution would likely be highly influential or conceptually important.