Hopfield model thermodynamic limit, AI-generated proofs
Awaiting review
The takeaway
Adriano Barra and Fabrizio Durante (arXiv 2610.41938, 30 Oct 2026) prove that the quenched pressure of the Hopfield model with K ~ αN Rademacher patterns has a thermodynamic limit at every load, temperature and uniform field, with no replica-symmetry assumption.
Status
- Claim
Awaiting review
- Our reporting
- Medium confidence
- Verification
- Preprint only
- Importance
- 3 of 5
- Last verified
- 2 November 2026
Your AI and this story
- GPT-6 Astra183 days after its cutoff
- Claude Opus 5.5122 days after its cutoff
- Gemini 4 Argon213 days after its cutoff
- Grok 4.7152 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 122 days before it.
Key facts
- Result: the limit equals the supremum over r of limiting pressures of mixed models keeping r binary patterns in a Gaussian bulk; existence only, no explicit formula
- Also gives an alternative proof of the Gaussian-pattern existence result announced by Camilli, Contucci and Mingione, extended to a uniform field
- AI disclaimer (verbatim): ‘The proofs presented in this paper were generated through extensive interaction with the AI models Claude (Opus 5.5) and ChatGPT (Sol 6.1), and the authors do not claim credit for their derivation … This write-up is made available as a service to the research community and is not intended for journal publication in its current form. All statements and proofs have been checked by the authors’
- Verification status: preprint, checked by the authors; no Lean
What happened
Barra and Durante posted a proof that the Hopfield associative-memory model has a well-defined infinite-volume free energy at every storage load, a question open in rigorous spin-glass theory (Guerra–Toninelli settled the Sherrington–Kirkpatrick case in 2002). Their method extracts large-overlap patterns, interpolates the rest to Gaussians, and reduces to results of Ho and of Chen, Ho and Issa on vector spin glasses.
The paper’s disclaimer credits the proofs to Claude Opus 5.5 and ChatGPT Sol 6.1, with the authors posing questions and suggesting approaches.
Why it matters
It follows Panchenko’s Mézard–Parisi proof found by ChatGPT Astra in the same week, another rigorous statistical-mechanics result whose authors attribute the derivation to models. Adriano Barra is a known researcher on Hopfield networks.
What is disputed or not yet verified
| Verification | Preprint, author-checked |
|---|
Sources
1 source from 1 site. Numbers match the chips in the text.
1 source: 1 primary
Primary
Changes
- Filed from the 2 Nov sweep