List your company
ResearchFormerly WizWand

Medical Question Answering on MedMCQA (Pass@1)

53.6Pass@1 AccuracyQwen2.5-7B + PPO
40.475243.882647.2950.6974Sep 28, 2025Nov 7, 2025Dec 17, 2025Jan 26, 2026Mar 7, 2026Apr 16, 2026May 27, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen2.5-7B + PPOBase Model=Qwen2.5-7B, Training=PPO2025.0953.6
Qwen2.5-7B + PPO w/ VERL.Base Model=Qwen2.5-7B, Training=PPO w/ VERL.2025.0952.9
Qwen2.5-7B + GRPOBase Model=Qwen2.5-7B, Training=GRPO2025.0952.1
Qwen2.5-7B + GRPO w/ VERL.Base Model=Qwen2.5-7B, Training=GRPO w/ VERL.2025.0952
Qwen2.5-7BBase Model=Qwen2.5-7B2025.0950.6
Qwen3-1.7BTrain Size=100%2026.0548.54
SHIFTTrain Size=0.2%2026.0548.47
SHIFTTrain Size=0.1%2026.0548.3
SC-EntropyTrain Size=0.2%2026.0547.37
ClusterTrain Size=0.2%2026.0547.3
Mathstral-7B-v0.1 + GRPO w/ VERL.Base Model=Mathstral-7B-v0.1, Training=GRPO w/ VERL.2025.0947.1
Random (avg)Train Size=0.2%2026.0546.97
A-PPLTrain Size=0.2%2026.0546.52
CoreSetTrain Size=0.2%2026.0546.48
SC-EntropyTrain Size=0.1%2026.0546.38
CoT SimilarityTrain Size=0.1%2026.0546.09
Random (avg)Train Size=0.1%2026.0545.91
ClusterTrain Size=0.1%2026.0545.88
Q-PPLTrain Size=0.2%2026.0545.88
CoT SimilarityTrain Size=0.2%2026.0545.85
Q-PPLTrain Size=0.1%2026.0545.53
A-PPLTrain Size=0.1%2026.0545.17
Mathstral-7B-v0.1 + GRPOBase Model=Mathstral-7B-v0.1, Training=GRPO2025.0944.9
CoreSetTrain Size=0.1%2026.0544.89
Mathstral-7B-v0.1 + PPO w/ VERL.Base Model=Mathstral-7B-v0.1, Training=PPO w/ VERL.2025.0944.8
Mathstral-7B-v0.1 + PPOBase Model=Mathstral-7B-v0.1, Training=PPO2025.0944.2
Mathstral-7B-v0.1Base Model=Mathstral-7B-v0.12025.0942.9
Qwen3-1.7BTrain Size=N/A2026.0540.98