Overall Capability Merging on Multi-Task Average
56.56Avg AccuracySWUDI
Evaluation results
| Method | Links | |
|---|---|---|
| SWUDImode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 56.56 | |
| SWUDI-Amode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 56.33 | |
| Mixture Trainingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 56.22 | |
| OptMergemode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 56.18 | |
| WUDI Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 56 | |
| TIES Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 55.87 | |
| TIES w/ DAREmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 55.85 | |
| Task Arithmeticmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 55.35 | |
| TA w/ DAREmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 55.17 | |
| Iso-Cmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 53.8 | |
| InternVL2.5-Instructmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 53.48 | |
| TSV Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 52.5 | |
| Weight Averagemode=Full fine-tuning, backbone=InternVL2.5-1B2026.06 | 48.33 |