Search on HR-MM Search (Pass@4)
42Pass@4AXPO
Evaluation Results
| Method | Links | |
|---|---|---|
| AXPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 42 | |
| SFT + GRPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 38 | |
| BaseModel Size=32B, Backbone=Qwen3-VL-Thinking2026.05 | 38 | |
| SFTModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 36.4 | |
| SFTModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 35.7 | |
| GRPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 34.4 | |
| BaseModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 33.8 | |
| AXPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 33.4 | |
| SFT + GRPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 33.1 | |
| AXPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 32.1 | |
| GRPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 29.5 | |
| SFTModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 28.5 | |
| SFT + GRPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 28.2 | |
| BaseModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 27.5 | |
| GRPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 22 | |
| BaseModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 18.7 |