Resolves YES if a flaw is found, even if OpenAI fixes it. Bundled answers resolve YES if any one component fails. A gap between a Lean statement and its paper does not count as a flaw, the flaw must be in the correctness of the proof. Timezone AoE if that matters. Flaw claim should be acknowledged by OpenAI, or stand unrebutted for about two weeks. Unless otherwise specified, flaw should be in the central result, or anything in the chain that links to the final result for linked results like 034/033/036. Flaw must be nontrivial, tiny mistakes that can be fixed with one line do not count, like a number being 3 instead of 4.
I will not bet after setting probabilities.
Update 2026-10-08 (PST) (AI summary of creator comment): * For the option Rational Hodge conjecture for CM abelian varieties (032), the market refers specifically to the central CM abelian result, rather than the broader Hodge family of papers. It will not resolve YES solely due to the withdrawal of other papers in the family.
1,000@vi several of the papers in the Hodge family have now been withdrawn, but the central CM abelian result has not. Is this option about the family in general (in which case I think it should resolve yes already) or is it only about the central CM abelian result?
@eapache I meant it to be the central CM abelian result. Clarified: Unless otherwise specified, flaw should be in the central result, or anything in the chain that links to the final result for linked results like 034/033/036. Flaw must be nontrivial, tiny mistakes that can be fixed with one line do not count, like a number being 3 instead of 4.