On Kahn's flow conjecture

Peter Keevash

For every increasing antipodal function h on the Boolean cube and every nonnegative allocation of each nonempty Fourier coefficient’s squared mass among its coordinates, the induced linear function flows upward to h₊². This is Kahn’s flow conjecture.

Sources for this record

What was already known

Friedgut, Kahn, Kalai and Keller formulated the flow conjecture in 2018. Chang, Liu and Liu had already proved the weaker Chvátal and Kleitman inequalities, whose result became an input here.

AI involvement AI-centered

Peter Keevash credits GPT-6 Astra with finding the proof within his suggested approach. Keevash simplified and rewrote it; Chang, Liu and Liu retain credit for their earlier theorem.

The model found the missing proof within a human-suggested approach and using a separate human theorem; central AI work is documented, but the source does not support little human mathematical input.

Importance Notable

Settles a structural strengthening of the intersecting-family inequalities.

Limitations

Keevash supplied the approach and used the separate Chang–Liu–Liu theorem. He rewrote and explained the model-found proof, but no completed independent or machine check is documented.

Technical verification Reported

Keevash supplies a detailed rewritten proof but does not expressly report a completed independent check of the model-origin argument or a machine verification. Reported is the missing-check-evidence default.

Human understanding Full

Keevash explains the proof strategy, its dependence on the sharp correlation inequality, and the flow deduction in the human-rewritten manuscript.

Correction and publication history