Forensic analysis report
VF-9E41C7A2
Sample — simulated result
Sample document with representative values. Detector scores are illustrative; thresholds and detection-power figures come from calibration run full_v4.
Exhibit
- Filename
- exhibit_2216-042.jpg
- Type
- image/jpeg
- Dimensions
- 1280 × 960
- Size
- 280.9 KB
9e41c7a2d05b8ff31c26e94a7d10b5c8e2f6a03d914b7ce50a8d21f6b3e49c77Calibrated decision
0.941
primary detector score · 0 = real · 1 = AI
The primary score exceeds the strict threshold 0.725, calibrated under Social chain to at most 1% false positives on every real source of run full_v4. Calibrated detection rate at this standard under Social chain: 43%.
thresholds calibrated for the Social chain condition — never a fixed 0.5
Primary detector · drives the verdict
Community Forensics — VF fine-tune
CVPR 2025 base · ViT-S/16 · 384px · pinned checkpoint
0.941
Supporting signals · corroboration only, never decide
Community Forensics (stock)
CVPR 2025 · ViT-S/16 · 384px
0.212
SDXL Detector
Community (HF) · Swin transformer classifier
0.716
a low supporting score does not weaken the verdict — stock models are known to lag on 2024+ generators
Robustness · detection power by condition
share of AI images caught at ≤5% false positives · run full_v4
honesty gap 0.504
Headline ranking quality (AUC 0.933) minus the calibrated detection rate at your chosen standard (43%). The share of benchmark performance that does not survive the evidentiary threshold — reported, not hidden.
Limits · read before acting on this report
- 01A flag is a calibrated statistical signal, not proof of synthesis. Corroborate with provenance and context before acting.
- 02The image shows a social_chain fingerprint (screenshot of a re-encoded image); sensor-level signals are not recoverable at this condition.
- 03The stock-model score (0.212) is expected to lag on 2024+ generators — that gap is why the fine-tuned primary drives the verdict. Supporting signals corroborate; they never decide.
- 04At the strict ≤1% false-positive standard, the calibrated detection rate is 43% — absence of a flag is weak evidence of authenticity.
- 05Generators released after the calibration run may not be represented; scores on novel generators can be arbitrarily wrong.
- 06This verdict applies to the exact bytes hashed above. Any re-export or edit produces a different exhibit.
Chain of custody
9e41c7a2d05b8ff31c26e94a7d10b5c8e2f6a03d914b7ce50a8d21f6b3e49c774c8d1a96f0b3e7255d9c02aa8f16de433b7a90c1e5f2481706bd3c5a92e8f014