VibeMathedMath problems solved by AI

Courtade and Kumar's Coordinate-wise Mutual Information Question

The Courtade-Kumar conjecture (2014) posits that dictatorship functions maximize mutual information between a Boolean function's output and a noisy input. The paper resolves an open question posed by Courtade and Kumar themselves - a sharp bound of 1H(α)1-H(\alpha) on the sum of coordinate-wise mutual informations for arbitrary bias - and extends the proven high-noise range of the main conjecture via optimal entropy bounds.

Result
Proved(see note)
Status
Partial result
AI contribution
AI co-developed
Method
Argument
Field
Boolean functions; information theory
Posed by
Thomas Courtade, Gowtham Kumar
Year posed
2014
Years open
12y
Solved
2026-01-14
Model
Gemini Deep Think (larger internal version)
Vendor
Google DeepMind
Collaborators
Adel Javanmard, David P. Woodruff
Verification
Unreviewed
Publication
Preprint
Significance
22 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

Fully resolves the posed coordinate-wise question; the main Courtade-Kumar conjecture itself remains open outside the extended high-noise range.

What the AI did

"The results in this paper were obtained with significant interaction with a larger version of Google's Deep Think Gemini-based model. The authors verified the entire paper and take full responsibility." The acknowledgments thank the Deep Think team by name.

Source

arXiv

Discussion