Multilingual Commonsense Reasoning on XStoryCloze
70.4Accuracy (en)Pythia-6.9B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Pythia-6.9BInitialization=Vanilla, Zero-shot=true2025.06 | 70.4 | 53.9 | 50.3 | 53.8 | — | |
| TokAlign + LATBackbone=Pythia-6.9B, Initialization=TokAlign, Language Adaptation Tuning (+LAT)=true, Zero-shot=true2025.06 | 70.2 | 54.4 | 49.3 | 53.9 | — | |
| TokAlignBackbone=Pythia-6.9B, Initialization=TokAlign, Zero-shot=true2025.06 | 66.8 | 48.6 | 47.7 | 50 | — | |
| Pythia-1BInitialization=Vanilla, Zero-shot=true2025.06 | 64.3 | 48.6 | 48 | 52.9 | — | |
| TokAlign + LATBackbone=Pythia-1B, Initialization=TokAlign, Language Adaptation Tuning (+LAT)=true, Zero-shot=true2025.06 | 64 | 51.1 | 47.8 | 53.5 | — | |
| TokAlignBackbone=Pythia-1B, Initialization=TokAlign, Zero-shot=true2025.06 | 61 | 47.6 | 47.1 | 51 | — | |
| Focus + LATBackbone=Pythia-6.9B, Initialization=Focus, Language Adaptation Tuning (+LAT)=true, Zero-shot=true2025.06 | 61 | 48.7 | 47.7 | 53.7 | — | |
| ZeTT + LATBackbone=Pythia-1B, Initialization=ZeTT, Language Adaptation Tuning (+LAT)=true, Zero-shot=true2025.06 | 60.3 | 49.3 | 47.2 | 52.1 | — | |
| Focus + LATBackbone=Pythia-1B, Initialization=Focus, Language Adaptation Tuning (+LAT)=true, Zero-shot=true2025.06 | 55.8 | 48.8 | 47.6 | 50.4 | — | |
| ZeTTBackbone=Pythia-1B, Initialization=ZeTT, Zero-shot=true2025.06 | 53.2 | 46.9 | 46.9 | 48.1 | — | |
| FocusBackbone=Pythia-1B, Initialization=Focus, Zero-shot=true2025.06 | 45.9 | 47.8 | 48.8 | 46.5 | — | |
| FocusBackbone=Pythia-6.9B, Initialization=Focus, Zero-shot=true2025.06 | 44.5 | 47.1 | 47.8 | 47.1 | — | |
| EuroLMscenario=zero-shot2025.12 | — | — | — | — | 56.5 | |
| Gamayunscenario=zero-shot2025.12 | — | — | — | — | 60.5 | |
| Gemma3scenario=zero-shot2025.12 | — | — | — | — | 56.3 | |
| Llama3.2scenario=zero-shot2025.12 | — | — | — | — | 50.5 | |
| Qwen2.5scenario=zero-shot2025.12 | — | — | — | — | 55.2 | |
| Qwen3scenario=zero-shot, thinking_mode=no-thinking2025.12 | — | — | — | — | 52.4 |