Style Alignment Correlation Analysis on Chinese legal text stylistic variation dataset (test)
0.8271Pearson Correlation (r)CLASE-Mix
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CLASE-MixVariant=Hybrid fusion2026.02 | 0.8271 | 0.8109 | 0.618 | |
| CLASE-ObjVariant=Objective-only2026.02 | 0.7923 | 0.7568 | 0.5692 | |
| ROUGEType=Traditional n-gram, Reference-based=true2026.02 | 0.425 | 0.4807 | 0.3355 | |
| BLEUType=Traditional n-gram, Reference-based=true2026.02 | 0.416 | 0.3383 | 0.2306 | |
| F1Type=Character-level exact matching, Reference-based=true2026.02 | 0.3907 | 0.3214 | 0.2213 | |
| METEORType=Traditional n-gram, Reference-based=true2026.02 | 0.3586 | 0.2949 | 0.202 | |
| GPT-4o-miniJudge=GPT-4o-mini, Reference-based=true2026.02 | 0.3206 | 0.3061 | 0.2083 | |
| CLASE-SubjVariant=Subjective-only2026.02 | 0.3165 | 0.3063 | 0.2148 | |
| BERTScoreType=Embedding-based, Reference-based=true2026.02 | 0.3145 | 0.201 | 0.1378 | |
| Qwen2.5-72BJudge=Qwen-2.5-72B, Reference-based=true2026.02 | 0.3049 | 0.2482 | 0.1808 | |
| Qwen2.5-72BJudge=Qwen-2.5-72B, Reference-based=false2026.02 | 0.2416 | 0.2336 | 0.1855 | |
| GPT-4o-miniJudge=GPT-4o-mini, Reference-based=false2026.02 | 0.0689 | 0.0747 | 0.0528 |