Vision-Language Modeling on OCRZH
26AccuracyZPPO
Evaluation Results
| Method | Links | |
|---|---|---|
| ZPPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 26 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 25.2 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 24.9 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 24.5 | |
| Qwen3.5Model Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 24 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 23.9 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 23.7 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 23 | |
| ZPPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 18.7 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 17.5 | |
| Qwen3.5Model Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 17 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 17 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 16.7 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 16.7 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 16.2 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 15.9 |