Cross-lingual Cultural Consistency on BLEnD Non-Indo-European
0.022Max SigmaVanilla
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VanillaModel=Qwen-2.5-3B2026.04 | 0.022 | 0.355 | — | — | — | |
| Few-shotModel=Qwen-2.5-3B2026.04 | 0.022 | 0.392 | — | — | — | |
| SteeringModel=Qwen-2.5-3B2026.04 | 0.022 | — | 0.346 | 0.353 | 0.36 | |
| PersonaModel=Qwen-2.5-3B2026.04 | 0.022 | — | 0.352 | 0.378 | 0.404 | |
| C-3POModel=Qwen-2.5-3B2026.04 | 0.022 | 0.455 | — | — | — | |
| VanillaModel=Gemma-2-27B2026.04 | 0.022 | 0.54 | — | — | — | |
| Few-shotModel=Gemma-2-27B2026.04 | 0.022 | 0.551 | — | — | — | |
| SteeringModel=Gemma-2-27B2026.04 | 0.022 | — | 0.549 | 0.555 | 0.563 | |
| PersonaModel=Gemma-2-27B2026.04 | 0.022 | — | 0.558 | 0.584 | 0.622 | |
| C-3POModel=Gemma-2-27B2026.04 | 0.022 | 0.643 | — | — | — | |
| VanillaModel=Llama-3.1-8B2026.04 | 0.023 | 0.387 | — | — | — | |
| Few-shotModel=Llama-3.1-8B2026.04 | 0.023 | 0.426 | — | — | — | |
| SteeringModel=Llama-3.1-8B2026.04 | 0.023 | — | 0.377 | 0.383 | 0.386 | |
| PersonaModel=Llama-3.1-8B2026.04 | 0.023 | — | 0.351 | 0.385 | 0.443 | |
| C-3POModel=Llama-3.1-8B2026.04 | 0.023 | 0.458 | — | — | — |