Visual Emotion Recognition on VECBench In-Domain ID VER (test)
69.7FI-8 AccuracyEmoCaliber
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| EmoCaliberModel Type=Emotion-Oriented MLLM2025.12 | 69.7 | 70 | 54 | 52.81 | 28.3 | 24.28 | 68.1 | 67.72 | 55.03 | 53.7 | |
| EmoVITModel Type=Emotion-Oriented MLLM2025.12 | 59.7 | 58.16 | 44.4 | 48.34 | 15 | 15.12 | 59.4 | 59.17 | 44.63 | 45.2 | |
| Qwen3-VLModel Type=Open-Source MLLM2025.12 | 56.5 | 54.57 | 49.7 | 49.56 | 21.1 | 21.72 | 52.6 | 50.64 | 44.98 | 44.12 | |
| GPT-5Model Type=Proprietary MLLM2025.12 | 53.8 | 53.06 | 44.5 | 45.61 | 19.5 | 19.37 | 54.8 | 52.94 | 43.15 | 42.75 | |
| InternVL3.5Model Type=Open-Source MLLM2025.12 | 53.1 | 51.96 | 42 | 41.09 | 16.1 | 15.46 | 51.9 | 50.47 | 40.78 | 39.75 | |
| Qwen2.5-VLModel Type=Open-Source MLLM2025.12 | 51.9 | 50.95 | 43.6 | 43.87 | 21 | 21.11 | 47.3 | 45.01 | 40.73 | 40.24 | |
| Emotion-LLaMAModel Type=Emotion-Oriented MLLM2025.12 | 23.5 | 18.97 | 18.7 | 20.11 | 10.3 | 10.04 | 18.7 | 13.26 | 17.8 | 15.6 |