Open-ended generation on Vicuna
99.1Skywork Reward V2 ScoreStudent
Evaluation Results
| Method | Links | |
|---|---|---|
| StudentBase Model=Gemma-3-4B2026.04 | 99.1 | |
| DistillableBase Model=Gemma-3-12B2026.04 | 99 | |
| TeacherBase Model=Gemma-3-12B2026.04 | 98.9 | |
| UndistillableBase Model=Gemma-3-12B2026.04 | 98.9 | |
| TeacherBase Model=Qwen3-8B2026.04 | 96.7 | |
| DistillableBase Model=Qwen3-8B2026.04 | 96.6 | |
| UndistillableBase Model=Qwen3-8B2026.04 | 96.5 | |
| ImproveBase Model=Gemma-3-4B2026.04 | 90.2 | |
| StudentBase Model=Qwen3-1.7B2026.04 | 89.1 | |
| ImproveBase Model=Qwen3-1.7B2026.04 | 87.4 | |
| GKD-FKLBase Model=Qwen3-1.7B2026.04 | 86.3 | |
| GKD-FKLBase Model=Gemma-3-4B2026.04 | 85.7 | |
| SFTBase Model=Gemma-3-4B2026.04 | 82.1 | |
| GKD-RKLBase Model=Gemma-3-4B2026.04 | 81.5 | |
| SFTBase Model=Qwen3-1.7B2026.04 | 81.4 | |
| GKD-RKLBase Model=Qwen3-1.7B2026.04 | 76.4 | |
| MisledBase Model=Gemma-3-4B2026.04 | 32.9 | |
| MisledBase Model=Qwen3-1.7B2026.04 | 31.2 |