Supplied
Labels, predictions and metadata from the file; provenance and label correctness are not verified.
FROM PREDICTIONS TO EVIDENCE
What does this result establish?
Compare precomputed predictions with reference labels. Inspect errors and record what remains unverified.
Binary classification only. No AI inference, trust score or deployment safety certification.
Loading teaching sample…
1 · Evaluation scope
Provenance and content identity
Metadata is declared by the importer and is not authenticated. SHA-256 identifies normalized input content; it does not establish trustworthy provenance.
2 · Results within the imported rows
N/A means a zero denominator, not 0%. These are rates within the imported data, not probabilities of correct predictions in real use. F1 = 2TP / (2TP + FP + FN).
3 · Inspect individual cases
| ID | Reference | Prediction | Outcome |
|---|
Class 1 does not automatically mean an attack. Read its definition above. These rows are not mapped to vehicles or timestamps in the 3D scene.
4 · Evidence and open questions
Derived
TP/TN/FP/FN and metrics computed from these rows only.
Assumed
Reference labels, data splits and threshold policy correctly describe the imported results.
Unknown
Leakage, generalization, calibration, robustness and deployment suitability.
Export a report to keep your notes. Loading another dataset clears existing notes; you will be asked first if notes are present.