Chowla conjecture
Let be the Liouville function. For every pair of distinct fixed nonnegative integers , one has .
References
Progress summary
An unverified OpenAI preprint is claimed to prove the conjecture, whereas earlier work established only weaker averaged versions.
The conjecture asserts cancellation in the ordinary, unweighted two-point correlation of Liouville values at every pair of distinct fixed shifts. Earlier literature established logarithmically averaged analogues, but not this ordinary limit.
Known results
- Tao (2015): proved the logarithmically averaged two-point result for distinct affine shifts.
- Helfgott and Radziwiłł (2022): obtained quantitative improvements for the logarithmic formulation.
- Pilatte (2023): improved logarithmic savings for two-point correlations.
- Matomäki, Radziwiłł, and Tao (2026): obtained further bounds for averaged Chowla formulations.
October 7, 2026 claimed solution
A preprint in an OpenAI repository is claimed to prove the full ordinary two-point conjecture with quantitative savings. A specialist commentary describes the claim as using existing methods, but the result has not undergone independent verification; the claim therefore remains unverified.
Community submission (unverified; October 7, 2026)
A submitted claim identifies an OpenAI preprint on ordinary two-point correlations as addressing the conjecture; this submission itself does not establish correctness.
Current status (as of October 2026): an OpenAI preprint is claimed to settle the ordinary two-point Chowla conjecture, but that claim is unverified; logarithmically averaged and other weaker forms are established.
Solutions 1
Claimed by OpenAI: ordinary two-point Chowla and corrected binary Elliott correlations.See full solution
Claimed by OpenAI. The attached preprint concerns ordinary two-point correlations of multiplicative functions, including the ordinary two-point Chowla conjecture and corrected binary Elliott correlations.
GitHub: https://github.com/openai/math
- 007.pdfOpen