Translator-call Mode Selection on PolyMath Low
87.64Macro F1 ScoreLUAR
Evaluation Results
| Method | Links | |
|---|---|---|
| LUARBackbone=Qwen3-4B2026.06 | 87.64 | |
| ST(qr)Backbone=Qwen3-8B2026.06 | 82.92 | |
| ST(qr)Backbone=Qwen3-4B2026.06 | 79.26 | |
| LUARBackbone=Qwen3-8B2026.06 | 78.32 | |
| BOUNDARY-SFTBackbone=Qwen3-4B2026.06 | 73.14 | |
| BOUNDARY-SFTBackbone=Qwen3-8B2026.06 | 69.2 | |
| ST(q)Backbone=Qwen3-8B2026.06 | 64.01 | |
| ST(q)Backbone=Qwen3-4B2026.06 | 63.25 | |
| NATIVE-TOOL-USEBackbone=Qwen3-8B2026.06 | 58.41 | |
| NATIVE-TOOL-USEBackbone=Qwen3-4B2026.06 | 57.6 | |
| SELF-ASSESSMENTBackbone=Qwen3-8B2026.06 | 43.48 | |
| SELF-ASSESSMENTBackbone=Qwen3-4B2026.06 | 41.1 |