List your company
ResearchFormerly WizWand

Vision-Language Modeling on BabyV

18.6AccuracyZPPO
6.2249.43712.6515.863Jun 16, 2026
Updated 4mo ago

Evaluation results

MethodLinks
ZPPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0618.6
GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0614.4
ZPPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.0613.9
GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0613.7
On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0613.1
On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0612.8
Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.0612.5
Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.0611.9
Qwen3.5Model Scale=2B, Prompt Replay Buffer Augmentation=false2026.0611.6
GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.069.8
GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.068.6
On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.067.8
On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.067.5
Qwen3.5Model Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.066.7
Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.066.7
Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.066.7