Cross-lingual Alignment Correlation on m-MMLU (test)
0.9859Pearson CorrelationNASCA
Evaluation Results
| Method | Links | |
|---|---|---|
| NASCABackbone=Mistral 0.3 7B, Aggregation Strategy=last2025.07 | 0.9859 | |
| MEXABackbone=Mistral 0.3 7B, Aggregation Strategy=average2025.07 | 0.9855 | |
| NASCABackbone=Qwen 2.5 7B, Aggregation Strategy=average2025.07 | 0.9849 | |
| NASCABackbone=Mistral 0.3 7B, Aggregation Strategy=average2025.07 | 0.9846 | |
| NAVCABackbone=Mistral 0.3 7B, Aggregation Strategy=average2025.07 | 0.9842 | |
| NASCABackbone=Qwen 2.5 7B, Aggregation Strategy=last2025.07 | 0.9792 | |
| NAVCABackbone=Llama 2 7B, Aggregation Strategy=last2025.07 | 0.979 | |
| NAVCABackbone=Llama 3.2 3B, Aggregation Strategy=average2025.07 | 0.9787 | |
| NAVCABackbone=Mistral 0.3 7B, Aggregation Strategy=last2025.07 | 0.9783 | |
| NAVCABackbone=Llama 2 7B, Aggregation Strategy=average2025.07 | 0.9762 | |
| NASCABackbone=Mistral 0.3 7B, Aggregation Strategy=weighted2025.07 | 0.9757 | |
| MEXABackbone=Llama 3.1 8B, Aggregation Strategy=average2025.07 | 0.972 | |
| NAVCABackbone=Llama 3.2 3B, Aggregation Strategy=last2025.07 | 0.9719 | |
| NASCABackbone=GLM 4 9B, Aggregation Strategy=average2025.07 | 0.9717 | |
| NASCABackbone=Llama 3.2 3B, Aggregation Strategy=weighted2025.07 | 0.9711 | |
| MEXABackbone=Llama 3.2 3B, Aggregation Strategy=weighted2025.07 | 0.9708 | |
| MEXABackbone=Qwen 2.5 7B, Aggregation Strategy=weighted2025.07 | 0.9705 | |
| NASCABackbone=Llama 3.1 8B, Aggregation Strategy=average2025.07 | 0.9704 | |
| NAVCABackbone=Llama 3.1 8B, Aggregation Strategy=average2025.07 | 0.9702 | |
| NAVCABackbone=Llama 3 8B, Aggregation Strategy=average2025.07 | 0.97 | |
| NASCABackbone=Llama 3.1 8B, Aggregation Strategy=weighted2025.07 | 0.97 | |
| MEXABackbone=Mistral 0.3 7B, Aggregation Strategy=weighted2025.07 | 0.9698 | |
| NASCABackbone=Llama 3.2 3B, Aggregation Strategy=last2025.07 | 0.9697 | |
| NASCABackbone=Qwen 2.5 7B, Aggregation Strategy=weighted2025.07 | 0.9696 | |
| NASCABackbone=Llama 3 8B, Aggregation Strategy=average2025.07 | 0.9693 | |
| NASCABackbone=Qwen 2.5 14B, Aggregation Strategy=average2025.07 | 0.9683 | |
| NAVCABackbone=Mistral 0.3 7B, Aggregation Strategy=weighted2025.07 | 0.9679 | |
| MEXABackbone=Llama 3 8B, Aggregation Strategy=average2025.07 | 0.9678 | |
| NASCABackbone=Llama 3.2 3B, Aggregation Strategy=average2025.07 | 0.9678 | |
| NASCABackbone=Llama 3.1 8B, Aggregation Strategy=last2025.07 | 0.9675 | |
| NAVCABackbone=GLM 4 9B, Aggregation Strategy=average2025.07 | 0.9673 | |
| NASCABackbone=Llama 3 8B, Aggregation Strategy=last2025.07 | 0.9669 | |
| NASCABackbone=Llama 3 8B, Aggregation Strategy=weighted2025.07 | 0.9663 | |
| NASCABackbone=GLM 4 9B, Aggregation Strategy=last2025.07 | 0.9661 | |
| NAVCABackbone=Llama 3.2 3B, Aggregation Strategy=weighted2025.07 | 0.9652 | |
| MEXABackbone=Llama 3.1 8B, Aggregation Strategy=weighted2025.07 | 0.9638 | |
| MEXABackbone=Llama 3 8B, Aggregation Strategy=weighted2025.07 | 0.9622 | |
| NAVCABackbone=Llama 3 8B, Aggregation Strategy=last2025.07 | 0.9612 | |
| NAVCABackbone=Llama 3.1 8B, Aggregation Strategy=last2025.07 | 0.9611 | |
| MEXABackbone=GLM 4 9B, Aggregation Strategy=weighted2025.07 | 0.9599 | |
| NASCABackbone=Qwen 2.5 14B, Aggregation Strategy=last2025.07 | 0.9597 | |
| NASCABackbone=GLM 4 9B, Aggregation Strategy=weighted2025.07 | 0.9578 | |
| NAVCABackbone=GLM 4 9B, Aggregation Strategy=last2025.07 | 0.9578 | |
| NASCABackbone=Llama 2 7B, Aggregation Strategy=last2025.07 | 0.9574 | |
| NAVCABackbone=Qwen 2.5 7B, Aggregation Strategy=weighted2025.07 | 0.9557 | |
| MEXABackbone=Llama 3.2 3B, Aggregation Strategy=average2025.07 | 0.9543 | |
| NASCABackbone=Llama 2 7B, Aggregation Strategy=average2025.07 | 0.9541 | |
| NAVCABackbone=Llama 3.1 8B, Aggregation Strategy=weighted2025.07 | 0.9504 | |
| NAVCABackbone=Qwen 2.5 7B, Aggregation Strategy=average2025.07 | 0.9499 | |
| NAVCABackbone=Llama 3 8B, Aggregation Strategy=weighted2025.07 | 0.9433 | |
| NAVCABackbone=Qwen 2.5 7B, Aggregation Strategy=last2025.07 | 0.9351 | |
| NAVCABackbone=Qwen 2.5 14B, Aggregation Strategy=average2025.07 | 0.9322 | |
| MEXABackbone=Llama 2 7B, Aggregation Strategy=weighted2025.07 | 0.93 | |
| NAVCABackbone=GLM 4 9B, Aggregation Strategy=weighted2025.07 | 0.9277 | |
| MEXABackbone=Llama 2 7B, Aggregation Strategy=average2025.07 | 0.9232 | |
| MEXABackbone=Qwen 2.5 14B, Aggregation Strategy=weighted2025.07 | 0.917 | |
| NAVCABackbone=Qwen 2.5 14B, Aggregation Strategy=last2025.07 | 0.917 | |
| NASCABackbone=Qwen 2.5 14B, Aggregation Strategy=weighted2025.07 | 0.9086 | |
| MEXABackbone=OLMO 2 7B, Aggregation Strategy=weighted2025.07 | 0.9076 | |
| NASCABackbone=OLMO 2 7B, Aggregation Strategy=weighted2025.07 | 0.9035 | |
| MEXABackbone=GLM 4 9B, Aggregation Strategy=average2025.07 | 0.8873 | |
| NASCABackbone=OLMO 2 7B, Aggregation Strategy=average2025.07 | 0.8871 | |
| MEXABackbone=Llama 2 7B, Aggregation Strategy=last2025.07 | 0.8861 | |
| NASCABackbone=OLMO 2 7B, Aggregation Strategy=last2025.07 | 0.886 | |
| NAVCABackbone=OLMO 2 7B, Aggregation Strategy=weighted2025.07 | 0.8802 | |
| MEXABackbone=OLMO 2 7B, Aggregation Strategy=average2025.07 | 0.8797 | |
| MEXABackbone=OLMO 2 7B, Aggregation Strategy=last2025.07 | 0.8772 | |
| NAVCABackbone=OLMO 2 7B, Aggregation Strategy=average2025.07 | 0.8663 | |
| MEXABackbone=Mistral 0.3 7B, Aggregation Strategy=last2025.07 | 0.8614 | |
| MEXABackbone=Qwen 2.5 7B, Aggregation Strategy=average2025.07 | 0.856 | |
| NAVCABackbone=Qwen 2.5 14B, Aggregation Strategy=weighted2025.07 | 0.8539 | |
| NAVCABackbone=OLMO 2 7B, Aggregation Strategy=last2025.07 | 0.8504 | |
| MEXABackbone=Llama 3 8B, Aggregation Strategy=last2025.07 | 0.8471 | |
| MEXABackbone=Llama 3.2 3B, Aggregation Strategy=last2025.07 | 0.8448 | |
| MEXABackbone=Llama 3.1 8B, Aggregation Strategy=last2025.07 | 0.8443 | |
| NASCABackbone=Llama 2 7B, Aggregation Strategy=weighted2025.07 | 0.8347 | |
| MEXABackbone=GLM 4 9B, Aggregation Strategy=last2025.07 | 0.8156 | |
| NAVCABackbone=Llama 2 7B, Aggregation Strategy=weighted2025.07 | 0.8086 | |
| MEXABackbone=Qwen 2.5 14B, Aggregation Strategy=average2025.07 | 0.7293 | |
| MEXABackbone=Qwen 2.5 7B, Aggregation Strategy=last2025.07 | 0.6312 | |
| MEXABackbone=Qwen 2.5 14B, Aggregation Strategy=last2025.07 | 0.6226 |