Multimodal Consistency Evaluation on ConBench
8.41ECEGPT-4o
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4oConf.=Verbal Confidence, Source=Closed-source MLLMs2026.04 | 8.41 | 77.71 | 94.35 | 64.98 | |
| Qwen2.5-VL-7BConf.=Token Confidence, Source=Open-source MLLMs2026.04 | 9.32 | 85.72 | 95.29 | 78.86 | |
| Phi-3.5-VisionConf.=Token Confidence, Source=Open-source MLLMs2026.04 | 10.58 | 80.78 | 89.28 | 64.84 | |
| MiniCPM-V-2.6Conf.=Verbal Confidence, Source=Open-source MLLMs2026.04 | 10.98 | 62.96 | 82.27 | 28 | |
| InternVL2.5-4BConf.=Token Confidence, Source=Open-source MLLMs2026.04 | 12.26 | 81.84 | 95.11 | 74.24 | |
| GPT-4oConf.=Token Confidence, Source=Closed-source MLLMs2026.04 | 12.33 | 90.52 | 98.22 | 88.99 | |
| InternVL3.5-8BConf.=Verbal Confidence, Source=Open-source MLLMs2026.04 | 13.12 | 65.12 | 85.89 | 35.86 | |
| InternVL2.5-4BConf.=Verbal Confidence, Source=Open-source MLLMs2026.04 | 13.87 | 61.26 | 87.27 | 33.02 | |
| MiniCPM-V-2.6Conf.=Token Confidence, Source=Open-source MLLMs2026.04 | 14.08 | 70.19 | 89.19 | 56.1 | |
| GPT-4o-miniConf.=Verbal Confidence, Source=Closed-source MLLMs2026.04 | 14.1 | 64.82 | 83.41 | 27.87 | |
| InternVL3.5-8BConf.=Token Confidence, Source=Open-source MLLMs2026.04 | 15.98 | 83.33 | 94.08 | 73.11 | |
| GPT-4o-miniConf.=Token Confidence, Source=Closed-source MLLMs2026.04 | 19.71 | 80.46 | 93.01 | 69.62 | |
| Phi-3.5-VisionConf.=Verbal Confidence, Source=Open-source MLLMs2026.04 | 20.33 | 63.14 | 72.65 | 10.32 | |
| Qwen2.5-VL-7BConf.=Verbal Confidence, Source=Open-source MLLMs2026.04 | 38.65 | 35.91 | 74.9 | -12.64 |