General Multilingual Capability on Aggregate 8 Seen Languages
69.5Average ScoreMGDA
Evaluation Results
| Method | Links | |
|---|---|---|
| MGDABackbone=Qwen3-8B-Base2026.06 | 69.5 | |
| CAGradBackbone=Qwen3-8B-Base2026.06 | 69.5 | |
| PCGradBackbone=Qwen3-8B-Base2026.06 | 69.2 | |
| BaseBackbone=Qwen3-8B-Base2026.06 | 68.4 | |
| V. SFTBackbone=Qwen3-8B-Base2026.06 | 67.6 | |
| CAGradBackbone=Qwen3-4B-Base2026.06 | 64 | |
| PCGradBackbone=Qwen3-4B-Base2026.06 | 64 | |
| MGDABackbone=Qwen3-4B-Base2026.06 | 63.9 | |
| BaseBackbone=Qwen3-4B-Base2026.06 | 63.4 | |
| V. SFTBackbone=Qwen3-4B-Base2026.06 | 62 | |
| CAGradBackbone=Llama-3.1-8B2026.06 | 47.3 | |
| MGDABackbone=Llama-3.1-8B2026.06 | 47.2 | |
| PCGradBackbone=Meta-Llama-3-8B2026.06 | 46.9 | |
| PCGradBackbone=Llama-3.1-8B2026.06 | 46.9 | |
| MGDABackbone=Meta-Llama-3-8B2026.06 | 46.8 | |
| CAGradBackbone=Meta-Llama-3-8B2026.06 | 46.2 | |
| V. SFTBackbone=Llama-3.1-8B2026.06 | 44.7 | |
| V. SFTBackbone=Meta-Llama-3-8B2026.06 | 44 | |
| BaseBackbone=Llama-3.1-8B2026.06 | 43.5 | |
| BaseBackbone=Meta-Llama-3-8B2026.06 | 38.8 |