List your company
ResearchFormerly WizWand

Vision-Language Understanding on MVista

80.5AccuracyZPPO
15.60432.45249.366.148May 19, 2026May 23, 2026May 28, 2026Jun 2, 2026Jun 6, 2026Jun 11, 2026Jun 16, 2026
Updated 4mo ago

Evaluation results

MethodLinks
ZPPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0680.5
GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0679.3
GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0679
Qwen3.5Model Scale=2B, Prompt Replay Buffer Augmentation=false2026.0678.6
On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0678.2
Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0677.9
On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0677.9
Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0677.8
Qwen3-VL-8BTraining Strategy=Staged2026.0575.9
Qwen3-VL-8BTraining Strategy=Merged2026.0573.8
ZPPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.0673.2
Qwen3-VL-8BTraining Strategy=Base2026.0572.4
Qwen2.5-VL-7BTraining Strategy=Staged2026.0571.45
InternVL3.5-8BTraining Strategy=Staged2026.0570.3
Qwen2.5-VL-7BTraining Strategy=Merged2026.0569.75
InternVL3.5-8BTraining Strategy=Merged2026.0569.4
Qwen2.5-VL-7BTraining Strategy=Base2026.0568.4
GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.0668.3
InternVL3-8BTraining Strategy=Staged2026.0565.4
GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.0665.2
On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.0663.6
On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.0662.7
Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.0662.2
Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.0662
InternVL3.5-8BTraining Strategy=Base2026.0560.7
InternVL3-8BTraining Strategy=Merged2026.0560.7
Qwen3.5Model Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.0660.7
InternVL3-8BTraining Strategy=Base2026.0518.1