List your company

Selective Prediction on To See or To Please Diagnostic

73.5Baseline AccuracyQwen2.5-VL-72B
66.63668.41870.271.982Mar 19, 2026

Evaluation results

MethodLinks
Qwen2.5-VL-72BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0373.577.84.3
Qwen2.5-VL-7BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0372.181.69.5
Molmo2-4BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0371.376.65.3
Llama-3.2-11BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0369.969.90
Phi-3.5-VisionStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.036969.70.7
LLaVA-NeXT-7BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0368.175.47.3
Pixtral-12BStrategy=Diagnostic-Guided Selective Prediction, Coverage=50%2026.0366.966.10.8