QED: An Open-Source Multi-Agent System for Generating Mathematical Proofs on Open Problems

Chenyang An, Qihao Ye, Minghao Pan, Jiayaun Zhang

For the discrete-time switch–walk–switch walk on Z₂ wr Z, started with all lamps off and walker positions 0 or 2, the total-variation distance between the time-t laws is comparable to t^(−1/2), with constants independent of t.

Sources for this record

What was already known

The report presents this sharp pairwise total-variation asymptotic as a research-level probability question supplied by a domain expert. It does not document a precise earlier bound for these two starting states; a larger priority or numerical improvement cannot be inferred.

AI involvement AI-autonomous

Minghao Pan supplied the problem statement. QED revised its proof plan and produced the proof without further human mathematical input; Pan then verified it.

The report documents problem-only human input and a model-driven proof search with substantial plan changes.

Importance Modest

Obtains the sharp rate for this specified pair of lamplighter starting states.

Limitations

This concerns two starting states for the line walk, not the separately published tree return-probability result. The reviewed report gives no precise earlier bound for these states or detailed proof-checking steps.

Technical verification Formal (human)

The QED report and project README say Minghao Pan verified the resulting proof of the exact line-walk total-variation asymptotic. This is a source-reported human check, not a local reproduction.

Formal checking

The QED report and project README say Minghao Pan verified the resulting proof of the exact line-walk total-variation asymptotic. This is a source-reported human check, not a local reproduction.

Checker
Minghao Pan, probability researcher at Caltech
Scope
||P_t^x-P_t^y||_TV asymptotic to within constants t^(-1/2) for the specified switch–walk–switch walk on Z_2 wr Z
Basis
Reported by the linked checker
Evidence source

Checking a formal statement does not establish that it matches the original problem.

Human understanding Few

The expert reports technical difficulty but gives no explanation of the central proof mechanism; Few is the evidence default, not a positive finding of incomprehension.

Correction and publication history