Perception on HRBen 8K (Pass@4)
90Pass@4AXPO
Evaluation Results
| Method | Links | |
|---|---|---|
| AXPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 90 | |
| BaseModel Size=32B, Backbone=Qwen3-VL-Thinking2026.05 | 90 | |
| AXPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 89 | |
| SFTModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 88 | |
| AXPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 88 | |
| SFTModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 88 | |
| SFTModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 87 | |
| SFT + GRPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 85.5 | |
| GRPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 85.5 | |
| SFT + GRPOModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 84.5 | |
| SFT + GRPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 84.5 | |
| BaseModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 82.5 | |
| GRPOModel Size=8B, Backbone=Qwen3-VL-Thinking2026.05 | 82 | |
| BaseModel Size=4B, Backbone=Qwen3-VL-Thinking2026.05 | 81 | |
| BaseModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 77 | |
| GRPOModel Size=2B, Backbone=Qwen3-VL-Thinking2026.05 | 72.5 |