Qwen2.5-VL-3B + SFT+DPO+GRPO
| Method | Links | |||
|---|---|---|---|---|
2026.03 | 2.73 | 2.62 | 2.93 | |
2026.03 | 2.73 | 2.42 | 2.75 | |
2026.03 | 2.73 | 2.7 | 2.94 | |
2026.03 | 2.71 | 2.64 | 2.89 | |
2026.03 | 2.7 | 2.13 | 2.42 | |
2026.03 | 2.7 | 2.65 | 2.93 | |
2026.03 | 2.64 | 2.24 | 2.33 | |
2026.03 | 2.62 | 2.17 | 2.55 | |
2026.03 | 2.6 | 1.61 | 2.01 | |
2026.03 | 2.55 | 2.53 | 2.93 | |
2026.03 | 2.54 | 1.6 | 1.92 | |
2026.03 | 2.5 | 2.28 | 2.84 | |
2026.03 | 2.42 | 2.34 | 2.8 | |
2026.03 | 2.34 | 2.46 | 2.08 | |
2026.03 | 2.31 | 2.29 | 2.73 | |
2026.03 | 2.1 | 2.17 | 1.97 | |
2026.03 | 2.08 | 2.35 | 2.19 | |
2026.03 | 1.33 | 2.31 | 2.02 | |
2026.03 | 1.31 | 2.31 | 1.98 | |
2026.03 | 0.97 | 1.76 | 1.37 | |
2026.03 | 0.4 | 2.25 | 1.62 | |
2026.03 | 0.05 | 2.37 | 1.7 |