Vision-Language Modeling on OCREN
48.8AccuracyZPPO
Evaluation results
| Method | Links | |
|---|---|---|
| ZPPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 48.8 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 46.4 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 46.1 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 45.5 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 45.3 | |
| Qwen3.5Model Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 44.7 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 44.5 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 44.4 | |
| ZPPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 42.5 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 41.3 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 40.9 | |
| Qwen3.5Model Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 40 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 39.6 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 39.5 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 39.1 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 38.9 |