Research · Papers · The SHA-256 record and exact synthesis · ML-046
On the miss rate of cubic rejection rules for exact K00 same-input H2 invalid points
Cubics cover the named H2 point and 57,794,560 exact K00 same-input H2 false points, missing 781,033,472 of 838,828,032
Published 2026-08-29
For everyone
Plain summary
ML-046 tests whether switching cubic rejection rules based on control bit k0 can catch the remaining invalid inputs in JSC-12. On the exact K00 same-input H2 test, the cubics miss 781,033,472 of 838,828,032 invalid points, or 93.1 percent. Four explicit misses land on the named case slice, alongside misses on JSC-13 and JSC-11. This closes the proposed K00 route. Exact counts for K01, K10, and K11 were not reported. The register records no prior art. Page verdict: SHOW.
Result
The cubics cover the named H2 point and 57,794,560 other exact K00 same-input H2 false points, but miss 781,033,472 of 838,828,032. Four explicit survivors land on the named case slice; JSC-13 and JSC-11 points are also missed.
Setting and definitions
K00, K01, K10, and K11 denote case labels. Control values k0=0 and k0=1 select the cubic pieces. The k0=0 zero set is {current1=0} ∪ {middle2=0} ∪ {σ₁+final1+final2=0}.
Method
An exact K00 false-language census and explicit-survivor report established the count. Sample screens of roughly 1–2k contexts across K01, K10, and K11 matched the ~93% miss rate, but their exact totals remain unenumerated.
Discussion
PROVED DEAD applies specifically to K00 same-input H2, not all K settings or alternative covers. Same-input fibres with current1=0 are invisible; the slice current1=1, σ₁=0 misses every H2 flip with middle2=0. Other K counts remain UNKNOWN. No score. Corrections: none. Prior-art position: unrecorded. Page verdict: SHOW.
For everyone — the takeaway
What this means
Testing two compact rules to catch invalid inputs showed they miss most of them. With 93% of K00 bad cases slipping through, these cubics cannot serve as a rejection filter. Samples suggest the same failure in other K settings, though full counts exist only for K00. The result rules out this specific design and gives future work a clear stop supported by concrete counterexamples.
Register references
- ML-046.
- Receipts: CONT
jsc_false_language_screen.json;ZKGOLF-FINDINGS-JSC-FALSE-LANGUAGE-2026-08-13.md. - Prior art: the register does not record this.
Every artifact named above is bundled in, or hashed by, this paper's evidence pack below.
Evidence pack
Everything needed to check this entry against its receipts: the register text, a manifest with a SHA-256 hash for every named receipt, and 2 of 2 receipt files bundled (11 KB). Anything not bundled is still hashed in the manifest and lives in the compute-box working trees.
Changelog
Last reviewed 2026-08-29
- 2026-08-29Published on this site.