VVDex ForgeAI exams. Evidence. Independent grading.

Direct measurement

Robot video annotation

meta/muse-spark-1.3 · meta/muse-spark-1.3 · Exam revision 0.2.0

3/8Certified passes · rate not published

Exact exam

vvdex.annotation.robot-video-1

Fingerprint
85e026007af17db692614084f12ff40c5bbf2348bcfda0a01deb95aad8928f4c
Revision status
Recorded revision

Input delivery

Video — native + exact frames

Evidence state
Proved
Adapter
vvdex-openrouter-native-video · 1
Provenance
Not established

Approved quality

Task-level measurements

Boxf1
0.75 · 8 samples
Boxiou
0.541137229944304 · 8 samples
Boxprecision
0.75 · 8 samples
Boxrecall
0.75 · 8 samples
Classificationaccuracy
1.0 · 8 samples
Classificationf1
1.0 · 8 samples
Classificationprecision
1.0 · 8 samples
Classificationrecall
1.0 · 8 samples
Eventf1
1.0 · 8 samples
Eventprecision
1.0 · 8 samples
Eventrecall
1.0 · 8 samples
Eventtemporaliou
0.7125000000000001 · 8 samples
Segmentf1
1.0 · 8 samples
Segmentprecision
1.0 · 8 samples
Segmentrecall
1.0 · 8 samples
Segmenttemporaliou
1.0 · 8 samples
Trackingconsistency
1.0 · 8 samples

Comparison conditions

Unproven

delivery adapter identity, version or configuration is not recorded; delivery conditions differ; delivery equivalence is unproven; delivery is not proved; different exam fingerprint; lane configuration is not recorded; repository cleanliness differs; execution equivalence is unproven; runtime limits differ

Evidence path

From measurement to sealed proof

Public evidence exposes commitments and approved report sections. Canonical task bodies, hidden tests, answers and private traces remain withheld.