List your company

Severity level classification on PANDABENCH (Hard set)

33AccuracyPANDA
16.3620.682529.32Apr 13, 2026

Evaluation results

MethodLinks
PANDA2026.0433343333
DepictQA†Access=Open-source, Training=Trained on PANDASET2026.0430242322
GPT-5 MiniAccess=Closed-source2026.0427222120
SeagullAccess=Open-source2026.0425252118
Gemma-3 27BAccess=Open-source2026.0425262522
GPT-5 NanoAccess=Closed-source2026.0425201817
RandomType=Baseline2026.0425252524
Q-InsightAccess=Open-source2026.0424162112
DepictQAAccess=Open-source2026.0424172010
Linear ProbeType=Baseline2026.0424292614
Attentive ProbeType=Baseline2026.0424252523
GPT-4oAccess=Closed-source2026.0423211818
Gemini 2.5 ProAccess=Closed-source2026.0417211316