Cross-lingual Cultural Consistency on BLEnD All 8 Languages
0.017Max SigmaVanilla
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VanillaModel=Qwen-2.5-3B2026.04 | 0.017 | 0.309 | — | — | — | |
| Few-shotModel=Qwen-2.5-3B2026.04 | 0.017 | 0.351 | — | — | — | |
| SteeringModel=Qwen-2.5-3B2026.04 | 0.017 | — | 0.304 | 0.309 | 0.311 | |
| PersonaModel=Qwen-2.5-3B2026.04 | 0.017 | — | 0.295 | 0.324 | 0.36 | |
| C-3POModel=Qwen-2.5-3B2026.04 | 0.017 | 0.405 | — | — | — | |
| VanillaModel=Llama-3.1-8B2026.04 | 0.019 | 0.409 | — | — | — | |
| Few-shotModel=Llama-3.1-8B2026.04 | 0.019 | 0.436 | — | — | — | |
| SteeringModel=Llama-3.1-8B2026.04 | 0.019 | — | 0.405 | 0.412 | 0.418 | |
| PersonaModel=Llama-3.1-8B2026.04 | 0.019 | — | 0.396 | 0.424 | 0.468 | |
| C-3POModel=Llama-3.1-8B2026.04 | 0.019 | 0.484 | — | — | — | |
| VanillaModel=Gemma-2-27B2026.04 | 0.019 | 0.581 | — | — | — | |
| Few-shotModel=Gemma-2-27B2026.04 | 0.019 | 0.573 | — | — | — | |
| SteeringModel=Gemma-2-27B2026.04 | 0.019 | — | 0.575 | 0.581 | 0.587 | |
| PersonaModel=Gemma-2-27B2026.04 | 0.019 | — | 0.593 | 0.618 | 0.642 | |
| C-3POModel=Gemma-2-27B2026.04 | 0.019 | 0.658 | — | — | — |