Search on HR-MM Search
25.9Pass@1SFT + AXPO
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.05 | 25.9 | |
| SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.05 | 24.4 | |
| SFTBackbone=Qwen3-VL-Thinking, Model Size=8B2026.05 | 23 | |
| BaseBackbone=Qwen3-VL-Thinking, Model Size=32B2026.05 | 22.8 | |
| BaseBackbone=Qwen3-VL-Thinking, Model Size=8B2026.05 | 21 | |
| SFTBackbone=Qwen3-VL-Thinking, Model Size=4B2026.05 | 20.6 | |
| GRPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.05 | 20.6 | |
| SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.05 | 20.1 | |
| SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.05 | 20 | |
| GRPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.05 | 18.4 | |
| SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.05 | 18.1 | |
| BaseBackbone=Qwen3-VL-Thinking, Model Size=4B2026.05 | 16.9 | |
| SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.05 | 15.7 | |
| SFTBackbone=Qwen3-VL-Thinking, Model Size=2B2026.05 | 14.5 | |
| GRPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.05 | 12.7 | |
| BaseBackbone=Qwen3-VL-Thinking, Model Size=2B2026.05 | 8.2 |