Vision-Language Modeling on VBlind
60.5AccuracyZPPO
Evaluation results
| Method | Links | |
|---|---|---|
| ZPPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 60.5 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 57 | |
| GRPOModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 56.5 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 55.6 | |
| Qwen3.5Model Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 55.2 | |
| On-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 54.9 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=true2026.06 | 54.6 | |
| Off-DistillModel Scale=2B, Prompt Replay Buffer Augmentation=false2026.06 | 53.7 | |
| ZPPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 44.7 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 43.6 | |
| GRPOModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 43.2 | |
| Qwen3.5Model Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 42.8 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 42.3 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=true2026.06 | 42.2 | |
| Off-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 41.8 | |
| On-DistillModel Scale=0.8B, Prompt Replay Buffer Augmentation=false2026.06 | 41.5 |