VibeMathedMath problems solved with AI

Improving Randomized Metric Distortion to 2.3282

In metric social choice, voters rank candidates by distance in an unknown metric space, while a randomized voting rule must use only these rankings. The paper introduces random-size stable lotteries and proves that, by mixing a suitably chosen random-size stable lottery with Integrated Veto, one obtains a randomized voting rule with metric distortion at most 11641/5000=2.328211641/5000=2.3282. This improves the previous best upper bound of 2.52.5. The proof combines infinite-dimensional conic linear-programming duality, heuristic nonlinear optimization, and exact rational verification using polynomial nonnegativity in the Bernstein basis.

Result
Proved(see note)
Status
Partial result
AI contribution
AI-discovered
Method
Argument
Field
Game theory
Posed by
Year posed
Years open
Solved
2026-08-29
Model
GPT-5.6 Sol; Claude Opus 5.0
Vendor
OpenAI; Anthropic
Collaborators
Nisarg Shah
Verification
Unreviewed
Publication
Preprint
Significance
18 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

The paper proves that there exists a randomized voting rule using only ordinal rankings with metric distortion at most 11641/5000=2.328211641/5000=2.3282. This improves the previous best upper bound of 2.52.5 and closes about 44%44\% of the gap to the known asymptotic lower bound of approximately 2.11262.1126. It does not determine the optimal randomized metric distortion: for m4m\ge4, the exact optimum and its asymptotic limit remain open.

What the AI did

GPT-5.6 Sol derived all mathematical proofs in the paper from research directions, literature connections, proof and search strategies, and inspiration supplied by Nisarg Shah. It autonomously introduced stable-lottery ingredients, developed progressively stronger bounds, and derived the proofs leading to 2.32822.3282. Shah then generalized one proposed lottery to random-size stable lotteries, guided the search over distributions, verified all final mathematical details, and rewrote and simplified the exposition with GPT-5.6 Sol and Claude Opus 5.

The disclosure is in the paper itself, not only in this entry: "All the proofs in this document were obtained using GPT-5.6-Sol with guidance from the author."

Verification

Unreviewed. The author states that he verified all final mathematical details; by this site's ladder an author's own check does not move the tier, however expert, and Shah is among the leading researchers on metric distortion. The bound 11641/500011641/5000 rests, per the abstract, on an exact rational verification via polynomial nonnegativity in the Bernstein basis, which is checkable in principle but has not been re-run here. No referee and no formalization.

Source

Submitted by VibeGene on

Changelog2 changes

Discussion