Content Moderation on Content Moderation Korean (test)
64.7Abusive RateCulturePark
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CultureParkBackbone=GPT-3.5-turbo2024.05 | 64.7 | 64 | 64.3 | |
| CultureBankTraining=fine-tuning GPT-3.5-turbo2024.05 | 63.5 | 52.2 | 57.9 | |
| Synatra-7B-v0.3-dpo2024.05 | 39 | 46.5 | 42.8 | |
| EEVE-Korean-10.8B-v1.02024.05 | 36.4 | 43.7 | 40 |