AI-detection and Quality Evaluation on AI-detection & Human-likeness Evaluation Chinese 1.0 (test)
69.4Agg ScoreHDPO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HDPOModel=Qwen2.5-1.5B2026.06 | 69.4 | 37.8 | 36.7 | 35.3 | 66.3 | |
| Bucket-SFTModel=Qwen2.5-1.5B2026.06 | 68.4 | 25.5 | 28.4 | 60 | 65.1 | |
| Bucket-SFTModel=Qwen2.5-3B2026.06 | 66.9 | 24.9 | 27.4 | 59.9 | 64.7 | |
| HDPOModel=Llama-3.2-3B2026.06 | 66.5 | 50.8 | 50.1 | 27.5 | 61.5 | |
| Bucket-SFTModel=Llama-3.2-3B2026.06 | 63.6 | 31.8 | 31.7 | 61.1 | 60.9 | |
| Full-SFTModel=Qwen2.5-1.5B2026.06 | 52.9 | 24.1 | 26.4 | 49.8 | 58.9 | |
| Full-SFTModel=Qwen2.5-3B2026.06 | 49.8 | 23.5 | 25.2 | 56.6 | 56.4 | |
| Full-SFTModel=Gemma-2-2B2026.06 | 48.9 | 22.7 | 27.5 | 54.7 | 55.8 | |
| Full-SFTModel=Llama-3.2-3B2026.06 | 43.3 | 32.5 | 34.2 | 43.8 | 51.4 | |
| BaseLMModel=Qwen2.5-3B2026.06 | 37.4 | 27.5 | 24 | 64.7 | 46.4 | |
| BaseLMModel=Qwen2.5-1.5B2026.06 | 29.3 | 28.1 | 27.2 | 58.1 | 39.8 | |
| BaseLMModel=Llama-3.2-3B2026.06 | 24.9 | 25.7 | 17.5 | 70.5 | 35.5 | |
| HDPOModel=Gemma-2-2B2026.06 | 24.5 | 40.8 | 39 | 35.6 | 35 | |
| Bucket-SFTModel=Gemma-2-2B2026.06 | 24.2 | 19.3 | 25 | 57.9 | 35.4 | |
| HDPOModel=Qwen2.5-3B2026.06 | 22.3 | 34.9 | 34 | 32.6 | 33.4 | |
| DPOModel=Qwen2.5-1.5B2026.06 | 18.4 | 40.8 | 38.3 | 59.9 | 27.4 | |
| DPOModel=Qwen2.5-3B2026.06 | 14.8 | 35 | 33.7 | 48.3 | 23.8 | |
| BaseLMModel=Gemma-2-2B2026.06 | 11.3 | 35.3 | 35.2 | 60 | 18.8 | |
| DPOModel=Llama-3.2-3B2026.06 | 8.8 | 38.3 | 46.6 | 46.9 | 15.2 | |
| DPOModel=Gemma-2-2B2026.06 | 5.8 | 37 | 43.9 | 34.9 | 10.6 |