List your company

Medical Visual Question Answering on VQA-RAD (BLEU, METEOR, CIDEr)

0.695BLEU-1Trajectory-Aware Process Supervision
0.37260.45630.540.6237Apr 1, 2026Apr 2, 2026Apr 4, 2026Apr 5, 2026Apr 7, 2026Apr 8, 2026Apr 10, 2026

Evaluation results

MethodLinks
Trajectory-Aware Process SupervisionReward Configuration=Rform + Rexact + RDTW2026.040.695———
GRPO (Rform + Rexact)Reward Configuration=Rform + Rexact2026.040.681———
SFTTraining Type=Supervised Fine-Tuning2026.040.597———
GRPO (Rform)Reward Configuration=Rform2026.040.571———
KG-CMI2026.040.4380.2680.2351.085
M3AE2026.040.4020.2350.210.925
MUMC2026.040.3850.2120.1950.852