Anari's Bethe permanent conjecture
For an nonnegative matrix , the Bethe permanent, which is computable in deterministic polynomial time, satisfies the tight universal comparisonThe lower bound, due to Gurvits, is attained on forests. The upper bound, due to Anari and Rezaei, is attained by the adjacency matrix of a disjoint union of -cycles. Confirming a conjecture of Anari, we provide an optimal girth-dependent refinement of the above comparison. More precisely, we show that if the bipartite support graph of has girth at least an even integer , thenThe upper bound is attained by the adjacency matrix of a disjoint union of -cycles.
- Result
- Proved(see note)
- Status
- Resolved
- AI contribution
- AI co-developed
- Method
- Argument
- Field
- Permanent approximation
- Posed by
- Nima Anari
- Year posed
- —
- Years open
- —
- Solved
- 2026-09-02
- Model
- ChatGPT 5.6 Sol Ultra
- Vendor
- OpenAI
- Collaborators
- Dingding Dong, Vishesh Jain
- Verification
- Unreviewed
- Publication
- Preprint
- Significance
- 22 / 100
- Disclosed cost
- —
- Wikipedia
- No dedicated article
What was actually shown
For every nonnegative matrix whose bipartite support graph has girth at least an even integer , Dong and Jain prove the sharp inequalityThe factor is optimal whenever , attained by matrices whose support graphs are disjoint unions of -cycles. Thus the result confirms Anari's conjecture, recovers the sharp universal Anari--Rezaei bound when , and approaches exactness as the support-graph girth tends to infinity.
What the AI did
The authors had already developed a strategy proving a weaker girth-dependent bound of the form , with . GPT-5.6 Sol Pro developed this strategy into a proof with . Subsequent interactions with GPT-5.6 Sol Ultra, focused on understanding the source of the logarithmic loss and its relation to the Anari--Rezaei argument, led to the development of the final optimal proof. Codex also assisted with manuscript preparation.
Verification
Unreviewed. A 21-page preprint two days old, no peer review, no formal verification. The optimality half is checkable by anyone: the paper exhibits the equality cases explicitly. The upper-bound argument is conventional and the authors take responsibility for it. Nobody independent has read it on the record.
Source
- PaperarXiv
Submitted by VibeGene on