Module probes 

2026-09-26: Capability probes: did the model see the picture? Each probe attaches fixtures, asks a question, and is scored by score::reply_matches against terms the reply must and must not contain. The fixtures they use come from scripts/gen_test_images.py, which draws a size label on each.

This leg shows that vision ran; geometry.rs checks the token count. Geometry alone would pass an encoder that produced the right count from wrong embeddings.

Owner: bench, vision. Invariants: none beyond the types.

Structs

Probe

2026-09-26: A probe: which fixtures to attach, what to ask, what a correct answer must and must not contain.

Constants

CONTROL

2026-09-26: The no-image control: the size-label question with nothing attached. It passes when the reply does not contain 1280, the term reads-the-size-label requires; otherwise the capability leg is not evidence and the run is VACUOUS.

PROBES

Functions

concurrency_probe

2026-09-26: The probe the concurrency leg sends, reads-the-size-label, so a reply passes only by reading the label from the image.