Condition: Naive — models given the questions only, no expert guidance
Figures shown: the BEST- and WORST-scoring single run per model × question — not a mean; ties on score are broken by answer value

Page generated: 2026-08-16
Data (model runs) generated: 2026-04-08
Scoring inputs: ground truth cc2ed8f5, evaluate.py 845b70e9

Open in GCS ↗

Image not available locally.

Open from GCS ↗

Run python analyze_page_viewer.py --download to fetch images.