What was already known
The general unentanglement oracle question was open; earlier approximate-disentangler lower bounds covered restricted error or model regimes.
AI involvement
The seven authors credit ChatGPT 5.6 Sol with the central proof idea after prompts to use the She–Yuen method. They combined, checked and developed the final argument.
The model supplied the central proof idea, while humans selected the She–Yuen starting point, merged accounts, verified and developed the final theorem and consequences.
Importance
Resolves the unitary-oracle separation question and gives an exponential input-size bound for approximate disentanglers when ε+δ<1.
Limitations
An oracle separation does not separate the unrelativized classes. A later classical-oracle result is separate work, and the authors’ verification of this proof was not independently repeated.
Technical verification
The authors explicitly say they verified the argument presented in the paper. This is a reported human check of the final proof; no independent rerun or machine formalization is documented.
Independent checking
author verification, simplification and development of the model-origin proof idea
- Checker
- John Bostanci, Sabee Grewal, Jonas Haferkamp, Andrew Huang, Yeongwoo Hwang, Anand Natarajan and Chinmay Nirkhe
- Scope
- main unitary-oracle separation argument and resulting no-disentanglers theorem
- Basis
- Reported by the linked checker
Human understanding
The authors explain how projectors with opposite product-state behavior reduce the invariant-polynomial lower bound to approximate degree of OR and derive the disentangler consequence.
Correction and publication history
This assessment covers the unitary-oracle separation; the later classical-oracle result is separate.
Read this published revision