FHERMA
BUILDJUDGINGPUBLICATION REQUIRED

Decompose an encrypted 170×256 image

sponsored by FairMath · 0 submissions

BOUNTYno bounty
JUDGINGJan 2026
Enter an implementation

ELIGIBILITY

ranking metricaccuracy
eligible runnersnot declared yet
input shape170 × 256
truncation rank30
schemeCKKS

MEASUREMENT PROTOCOL

  1. 01Correctness is validated before anything is timed. An entry that fails validation produces no result at all.
  2. 02Every submission is re-run on the same runner, so a ranking compares implementations rather than machines.
  3. 03The harness version and the runner instance are recorded with each execution and shown next to the number.
  4. 04Results stay attached to the specification version they were measured under; a later revision never rewrites them.

HARNESS

The benchmark harness belongs to the kernel, not to this challenge, and arrives with the measurement layer. Until then these rules describe intent rather than something the platform enforces.