Southeast Asian language understanding and generation on SEA-HELM
42Indonesian ScoreQwen2.5-3B-Instruct (Teacher)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Qwen2.5-3B-Instruct (Teacher)Model Type=Teacher2026.06 | 42 | 38.7 | 32.3 | 9.8 | 24.6 | 40 | 6.5 | 27.7 | |
| SPINTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 12.6 | 11.4 | 10 | 5.3 | 8 | 14.7 | 4.8 | 9.5 | |
| GKDTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 11.7 | 13.2 | 10.9 | 4.9 | 4.8 | 13.6 | 3.7 | 9 | |
| DuDiTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 11.7 | 14.4 | 12.8 | 6.5 | 6.1 | 14.8 | 4.6 | 10.1 | |
| SFTTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 10.6 | 11.8 | 10.8 | 4.9 | 5.6 | 12 | 4.4 | 8.6 | |
| DFTTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 9.3 | 11.6 | 10.2 | 10 | 6.1 | 12.4 | 8.3 | 9.7 | |
| SeqKDTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 7.1 | 9.1 | 7.7 | 4.2 | 3.4 | 8.1 | 2.9 | 6.1 | |
| SDFTTeacher Model=Qwen2.5-3B-Instruct, Student Model=Qwen2.5-0.5B2026.06 | 4.2 | 4.6 | 4.7 | 2.3 | 3.3 | 5.1 | 1.3 | 3.6 |