back to table · edit · history · where entries came from · files · download
1691 bytes, as of the version from 2026-11-01 17:38 (current). Recorded here, not run.
# Audit findings for T656
The authenticated site audit was run after filling. It returned `clean: false`
with exactly two advisory findings, retained verbatim in `audit.json`. Neither
is suppressed or represented as a clean audit.
1. The symbolic parameter `convention` has two labels, each repeated across
the dimensions. This is the intended one-number-in-two-conventions
exception: gamma_n=1/rho_n and rho_n=1/gamma_n. The definition explicitly
calls them invertible conventions of the unique minimizing argument; the
conventions comment explicitly says they are one number. The heading is
rho_n or gamma_n, not a generic value. The audit implementation asks this
structural question for every such selector; it does not inspect whether
the definition answers it. All 31 reciprocal identities were checked.
2. The existing tag `matrix multiplication` has public `table_count: 1`,
namely T5. The audit tests that cached public count against <=1. It does
not count these private drafts: T655 and T656. This tag is required by
the family conventions and connects three existing tables when the two
drafts are included. Both source papers discuss the connection of the
scalar family to reduced polynomial multiplication and restrictions on
matrix-multiplication constructions. The tag is retained rather than
creating a new tag or dropping a justified shared family classification.
T655 was read but not edited; it remains a draft at this build's handoff.
The independent critique must assess both explanations. This report is the
builder's response, not an independent acceptance. No offer, publication or
human-review action has been taken.