Reasoning on PolyMath 8 Seen Languages
50PolyMath ScoreMGDA
Evaluation Results
| Method | Links | |
|---|---|---|
| MGDABackbone=Qwen3-8B-Base2026.06 | 50 | |
| CAGradBackbone=Qwen3-8B-Base2026.06 | 49 | |
| PCGradBackbone=Qwen3-8B-Base2026.06 | 48.4 | |
| V. SFTBackbone=Qwen3-8B-Base2026.06 | 46.8 | |
| CAGradBackbone=Qwen3-4B-Base2026.06 | 44.9 | |
| MGDABackbone=Qwen3-4B-Base2026.06 | 44.7 | |
| PCGradBackbone=Qwen3-4B-Base2026.06 | 44.6 | |
| BaseBackbone=Qwen3-8B-Base2026.06 | 44.5 | |
| V. SFTBackbone=Qwen3-4B-Base2026.06 | 43.4 | |
| BaseBackbone=Qwen3-4B-Base2026.06 | 39.4 | |
| CAGradBackbone=Llama-3.1-8B2026.06 | 22.9 | |
| MGDABackbone=Llama-3.1-8B2026.06 | 22.4 | |
| PCGradBackbone=Llama-3.1-8B2026.06 | 22.4 | |
| MGDABackbone=Meta-Llama-3-8B2026.06 | 21.6 | |
| PCGradBackbone=Meta-Llama-3-8B2026.06 | 21.4 | |
| V. SFTBackbone=Llama-3.1-8B2026.06 | 20.3 | |
| CAGradBackbone=Meta-Llama-3-8B2026.06 | 20.2 | |
| V. SFTBackbone=Meta-Llama-3-8B2026.06 | 18.3 | |
| BaseBackbone=Llama-3.1-8B2026.06 | 13.2 | |
| BaseBackbone=Meta-Llama-3-8B2026.06 | 12 |