Instruction Following and Multimodal Reasoning on IFEval and MMBench
61.74IFEval ScoreQwen2.5-VL-7B-Instruct
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL-7B-InstructBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 61.74 | 84.02 | 72.88 | |
| TA + OrthoMerge-CBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 61 | 83.85 | 72.43 | |
| TA + OrthoMerge-GBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 60.81 | 83.94 | 72.38 | |
| TABackbone=Qwen2.5-VL-7B-Instruct2026.02 | 60.12 | 83.84 | 71.98 | |
| TSV-M + OrthoMerge-GBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 57.86 | 83.85 | 70.86 | |
| TIES + OrthoMerge-GBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 56.19 | 84.19 | 70.19 | |
| TIESBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 55.15 | 83.17 | 69.16 | |
| TIES + OrthoMerge-CBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 54.9 | 83.59 | 69.25 | |
| TSV-MBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 54.9 | 83.07 | 68.99 | |
| TSV-M + OrthoMerge-CBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 54.9 | 83.42 | 69.16 |