Short-form text rewriting on standardized evaluation set
4.94Semantic Fidelity ScoreGPT-5-Chat
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-5-Chat2026.05 | 4.94 | 4.79 | 4.49 | 3.13 | 4.98 | 22.33 | |
| Phi SilicaConfiguration=Finetuned2026.05 | 4.92 | 4.76 | 4.52 | 2.92 | 4.97 | 22.09 | |
| Phi SilicaConfiguration=Finetuned + Quantized2026.05 | 4.92 | 4.74 | 4.53 | 2.86 | 4.97 | 22.02 | |
| Phi SilicaConfiguration=Baseline2026.05 | 4.75 | 4.27 | 4.1 | 3.59 | 4.95 | 21.66 |