Image Captioning on MS-COCO 200 images (test)
1,020.38TokensLingoLoop
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LingoLoopMLLM=Qwen2.5-VL-3B2025.06 | 1,020.38 | 7,090.58 | 32.94 | |
| LingoLoopMLLM=InstructBLIP2025.06 | 1,002.08 | 3,152.26 | 57.3 | |
| LingoLoopMLLM=Qwen2.5-VL-7B2025.06 | 797.55 | 3,839.7 | 15.24 | |
| LingoLoopMLLM=InternVL3-8B2025.06 | 554.41 | 2,771.76 | 9.7 | |
| Verbose imagesMLLM=Qwen2.5-VL-3B2025.06 | 394.74 | 2,682.38 | 13.12 | |
| Verbose imagesMLLM=InternVL3-8B2025.06 | 362.38 | 1,810.23 | 6.4 | |
| Verbose imagesMLLM=Qwen2.5-VL-7B2025.06 | 345.59 | 1,738 | 6.99 | |
| Verbose imagesMLLM=InstructBLIP2025.06 | 332.29 | 1,241.89 | 17.79 | |
| NoneMLLM=Qwen2.5-VL-7B2025.06 | 88.86 | 445.25 | 1.84 | |
| NoiseMLLM=Qwen2.5-VL-7B2025.06 | 88.24 | 446.17 | 1.88 | |
| NoneMLLM=InstructBLIP2025.06 | 86.11 | 428.72 | 4.91 | |
| NoiseMLLM=InstructBLIP2025.06 | 85.78 | 426.22 | 4.95 | |
| NoneMLLM=InternVL3-8B2025.06 | 76.31 | 379.14 | 1.39 | |
| NoiseMLLM=InternVL3-8B2025.06 | 74.89 | 362.1 | 1.38 | |
| NoiseMLLM=Qwen2.5-VL-3B2025.06 | 68.07 | 440.25 | 2.4 | |
| NoneMLLM=Qwen2.5-VL-3B2025.06 | 66.64 | 430.01 | 2.24 |