Multimodal Summarization on XMSMO
4.08BLEU-4CoE
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CoEzero-shot=true, training-free=true, backbone=Qwen2.5-VL-7B-Instruct2026.03 | 4.08 | 21.28 | 56.55 | 17.24 | 88.65 | 33.82 | 3.65 | |
| TCoTzero-shot=true, training-free=true, backbone=Qwen2.5-VL-7B-Instruct2026.03 | 1.67 | 15.32 | 24.82 | 12.21 | 86.8 | 28.44 | 3.61 | |
| ViTCoTzero-shot=true, training-free=true, backbone=Qwen2.5-VL-7B-Instruct2026.03 | 1.67 | 13.41 | 10.61 | 14.47 | 86.5 | 27.86 | 3.81 | |
| CoFzero-shot=true, training-free=true, backbone=InternVL3-8B2026.03 | 0.51 | 8.64 | 1.6 | 8.8 | 80.77 | 27.67 | 3.03 | |
| CoSzero-shot=true, training-free=true, backbone=LLaVA-Video / LongVA2026.03 | 0.51 | 7.12 | 0.09 | 8.29 | 83.91 | 27.16 | 2.99 |