List your company

Medical Visual Question Answering on SLAKE-VQA (Sentence BLEU-1)

0.841Sentence BLEU-1Trajectory-Aware Process Supervision
0.741160.767080.7930.81892Apr 10, 2026

Evaluation results

MethodLinks
Trajectory-Aware Process SupervisionReward Configuration=Rform + Rexact + RDTW2026.040.841
GRPO (Rform + Rexact)Reward Configuration=Rform + Rexact2026.040.791
SFTTraining Type=Supervised Fine-Tuning2026.040.786
GRPO (Rform)Reward Configuration=Rform2026.040.745