VibeMathedMath problems solved with AI

Depth-1 distinctness for pseudorandom unitaries

A single layer of independent random single-qubit Clifford gates is negl(n)\operatorname{negl}(n)-distinct for polynomially many queries. Consequently, in the PFCPFC pseudorandom-unitary construction, the depth-logn\log n global unitary 22-design layer can be replaced by a depth-1 tensor product of single-qubit 22-designs while retaining the distinctness property needed for the construction. This provides a counterexample to the conjecture that every negligibly distinct ensemble must be entangling.

Result
Disproved(see note)
Status
Resolved
AI contribution
AI-discovered
Method
Construction
Field
Quantum pseudorandomness
Posed by
Raza, Eisert and Fefferman (the authors' own working conjecture)
Year posed
Years open
Solved
2026-09-02
Model
ChatGPT 5.6 Sol
Vendor
OpenAI
Collaborators
Asad Raza, Jens Eisert, Bill Fefferman
Verification
Unreviewed
Publication
Preprint
Significance
6 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

The paper proves that a tensor-product ensemble of independently chosen single-qubit Clifford gates is negl(n)\operatorname{negl}(n)-distinct when the number of queries is polynomial in nn. This disproves the authors' conjecture that negligibly distinct ensembles must necessarily be entangling. As an application, the depth-logn\log n global Clifford/unitary-22-design layer used in the PFCPFC pseudorandom-unitary construction can be replaced by a single depth-1 layer of local single-qubit 22-designs while preserving the required distinctness property.

What the AI did

The authors state that Proposition III.5 was proposed by ChatGPT 5.6 Sol as a counterexample to their conjecture that a negl(n)\operatorname{negl}(n)-distinct ensemble must be entangling. Sol identified that a single layer of independent random single-qubit Clifford gates is already negligibly distinct for polynomially many queries. ChatGPT 5.5 and 5.6 Pro were also used to devise proof strategies for the paper's main results, as well as for literature search, exposition, and technical checking; all proofs were independently verified by the authors.

Verification

The result appears as Proposition III.5 in a coauthored research paper by Asad Raza, Jens Eisert, and Bill Fefferman. The authors explicitly state that ChatGPT 5.6 Sol proposed the counterexample and that they independently verified all proofs. No formal proof-assistant certificate is reported.

Source

Submitted by VibeGene on

Changelog2 changes

Discussion