Utility on OPI clean
90.3Utility ScoreGemma-4-E4B-it
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemma-4-E4B-itSetting=Baseline2026.05 | 90.3 | |
| Gemma-4-E4B-itSetting=ACT2026.05 | 89.3 | |
| Gemma-4-E4B-itSetting=BCT2026.05 | 89.1 | |
| GPT-OSS-20BSetting=Baseline2026.05 | 88.8 | |
| GPT-OSS-20BSetting=BCT2026.05 | 86 | |
| Qwen3-8BSetting=Baseline2026.05 | 81.2 | |
| Qwen3-8BSetting=ACT2026.05 | 80.4 | |
| Qwen3-1.7BSetting=Baseline2026.05 | 80 | |
| Qwen3-1.7BSetting=BCT2026.05 | 78.2 | |
| Qwen3-8BSetting=BCT2026.05 | 78.2 | |
| GPT-OSS-20BSetting=ACT2026.05 | 72.1 | |
| Phi-4-reasoningSetting=Baseline2026.05 | 69.6 | |
| Qwen3-1.7BSetting=ACT2026.05 | 68.3 | |
| Phi-4-reasoningSetting=ACT2026.05 | 66.8 | |
| Phi-4-reasoningSetting=BCT2026.05 | 6.7 |