VibeMathedMath problems solved by AI

Pach's Tangency Conjecture: Improved Bounds

Pach conjectured that nn Jordan arcs, pairwise crossing exactly once with no triple points, have O(n)O(n) tangent pairs. The best known bound stood at O(n7/4)O(n^{7/4}); the paper improves it to O(n3/2)O(n^{3/2}) (and O(n5/3)O(n^{5/3}) in the at-most-one-crossing relaxation), plus a tight Θ(n4/3)\Theta(n^{4/3}) for a grounded x-monotone variant.

Result
Proved(see note)
Status
Partial result
AI contribution
AI-assisted
Method
Argument
Field
Combinatorial geometry
Posed by
János Pach
Year posed
Years open
Solved
2026-03-12
Model
Gemini
Vendor
Google DeepMind
Collaborators
Eyal Ackerman, Balázs Keszegh
Verification
Unreviewed
Publication
Preprint
Significance
15 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

Exponent improvements toward Pach's conjecture, which remains open.

What the AI did

"For the proof of Theorem 9 we used some back and forth interaction with Google's Large Language Model Gemini." One theorem of the paper, attributed plainly.

Source

arXiv

Discussion