Limit free energy of the Hopfield model identified at every temperature and loading ratio
Six mathematicians identified, by a Parisi-type variational formula, the limit of the free energy of the Hopfield associative-memory model with random Bernoulli patterns at every temperature and every ratio of patterns to neurons, after one of them obtained the existence of the limit with assistance from GPT 6 Astra.
- Model
- GPT 6 Astra (existence proof and writing) and Fable 5.1 (writing)
- Field
- Mathematics
- Date
- 2026-11-02
- Human collaborators
- Ayman Ali, Jeanne Boursier, Hong-Bin Chen, Fu-Hsuan Ho, Victor Issa, Jean-Christophe Mourrat
What was found
In the Hopfield model, spins interact through independent random patterns, and in the regime where the number of patterns grows in proportion to the number of spins, Amit, Gutfreund and Sompolinsky derived a non-rigorous phase diagram in the 1980s; rigorous results covered high temperature, small loading with low temperature, or assumed concentration of the order parameters. The paper proves that the expected free energy converges for every loading ratio and inverse temperature and identifies the limit as a sup-inf variational formula of Parisi type, with almost sure and convergence as well. A Hubbard-Stratonovich transformation turns the model into a bipartite spin glass with and spherical spins, which the authors show is equivalent to one with Gaussian patterns plus finitely many Bernoulli random fields, and they solve that model by adapting recent techniques for spin glasses. The statement of AI use says Ayman Ali obtained the existence proof with assistance from GPT 6 Astra and shared it, that Chen and Mourrat already had a detailed draft identifying the limit for Gaussian patterns, and that all six wrote the final manuscript with GPT 6 Astra and Fable 5.1.
Novelty check
arXiv, 2026-11-16, ti:Hopfield AND abs:"free energy" returned 21 records. The nearest are concurrent and cited by the paper: Camilli, Contucci and Mingione (2610.32618, 23 October 2026) solve the model with Gaussian patterns, and Barra and Durante (2610.41938, 30 October 2026) prove existence of the limit for Bernoulli patterns without identifying it. Earlier records cover the replica-symmetric regime only. Nothing in this registry covered the model.
Caveats and known objections
Author verified: the authors state that they manually verified all statements and proofs and take full responsibility, and no outside check exists yet for a 37-page preprint. Two concurrent papers within a fortnight cover parts of the result, so credit for its pieces is shared. Autonomy is ai-assisted: the model helped one author with the existence proof, while the identification of the limit, the main result, built on two of the authors' own draft for Gaussian patterns, and the rest of the model use was in writing.
Also recorded at
vibemathedfree-energy-of-the-hopfield-model
Nothing mechanically. It is a parallel listing of the same result, carrying its own verification label rather than an independent check.
Labels its own verification unreviewed.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Author verified and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded author verified for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Limit free energy of the Hopfield model identified at every temperature and loading ratio. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-02-hopfield-free-energy
BibTeX
@misc{whataifound-independent-2026-energy,
title = {Limit free energy of the Hopfield model identified at every temperature and loading ratio},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Independent (Hebrew University, MIT, NYU Shanghai, Weizmann Institute, Sorbonne University, ENS Lyon). Verification: Author verified. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-11-02-hopfield-free-energy}
}