Chinese hate speech detection on STATE-ToxiCN (test1)
35.53Overall ScorePrompt-Driven Large Language Model Merge
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Prompt-Driven Large Language Model MergeBase Model=Qwen2.5-7B-Instruct2025.12 | 35.53 | 25.04 | 46.04 | |
| SFTBase Model=Qwen2.5-7B-Instruct2025.12 | 34.36 | 23.83 | 44.89 | |
| CPT+SFTBase Model=Qwen2.5-7B-Base2025.12 | 33.79 | 23.53 | 44.04 | |
| RFTBase Model=Qwen2.5-3B-Instruct2025.12 | 20.21 | 11.26 | 29.15 |