List your company

Physical Reasoning Visual Question Answering on CRAFT VQA

39.4Descriptive Accuracy+EG+VAORA
26.08829.5443336.456Jul 7, 2026

Evaluation results

MethodLinks
+EG+VAORAbase model=Qwen3VL-8B-SFT, Expert-Guided reward=true, VAORA=true2026.0739.448.246.345.950.546.06
+EGbase model=Qwen3VL-8B-SFT, Expert-Guided reward=true2026.0738.544.54642.351.544.56
SFTbase model=Qwen3VL-8B2026.0737.746.943.944.747.544.14
Qwen3-VL-8B-Instruct2026.0726.641.740.74240.738.34