Cultural Alignment on WVS
70.44AccuracyGPT-5 + Role-Play
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5 + Role-PlayFamily=Proprietaty LLMs2025.04 | 70.44 | 8 | |
| Role-PlayFamily=GPT-4.12025.04 | 69.85 | 8 | |
| CAReDiOFamily=GPT-4.12025.04 | 69.66 | 8 | |
| CultureBankFamily=GPT-4.12025.04 | 68.11 | 8 | |
| CultureDataFamily=Gemma-3-27B-IT2025.04 | 68.02 | 6 | |
| CAReDiOFamily=Gemma-3-27B-IT2025.04 | 67.96 | 6 | |
| CultureSPAFamily=Gemma-3-27B-IT2025.04 | 67.76 | 6 | |
| Role-PlayFamily=Gemma-3-27B-IT2025.04 | 67.22 | 4 | |
| CultureBankFamily=Gemma-3-27B-IT2025.04 | 67.11 | 5 | |
| CultureLLMFamily=Gemma-3-27B-IT2025.04 | 66.99 | 4 | |
| CultureParkFamily=Gemma-3-27B-IT2025.04 | 65.95 | 4 | |
| CultureLLMFamily=Qwen2.5-7B-Instruct2025.04 | 65.49 | 7 | |
| CAReDiOFamily=Qwen2.5-7B-Instruct2025.04 | 65.26 | 8 | |
| Role-PlayFamily=Qwen2.5-7B-Instruct2025.04 | 64.96 | 8 | |
| Raw ModelFamily=Gemma-3-27B-IT2025.04 | 64.77 | — | |
| CultureDataFamily=Qwen2.5-7B-Instruct2025.04 | 64.69 | 8 | |
| CultureInstructionFamily=Qwen2.5-7B-Instruct2025.04 | 63.25 | 7 | |
| CultureBankFamily=Qwen2.5-7B-Instruct2025.04 | 62.47 | 7 | |
| GPT-5Family=Proprietaty LLMs2025.04 | 62.38 | — | |
| CultureSPAFamily=Qwen2.5-7B-Instruct2025.04 | 62 | 7 | |
| CultureInstructionFamily=Gemma-3-27B-IT2025.04 | 61.42 | 2 | |
| Raw ModelFamily=GPT-4.12025.04 | 60.91 | — | |
| Raw ModelFamily=Qwen2.5-7B-Instruct2025.04 | 59.68 | — | |
| CultureParkFamily=Qwen2.5-7B-Instruct2025.04 | 57.83 | 2 |