Visual Emotion Recognition on VECBench VER (OOD)
85.4UnbiasedEmo-6 AccuracyQwen3-VL
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Qwen3-VLCategory=Open-Source MLLM2025.12 | 85.4 | 85.47 | 23.58 | 22.87 | 40.32 | 41.1 | 49.77 | 49.81 | |
| EmoCaliberCategory=Emotion-Oriented MLLM2025.12 | 79.9 | 80.41 | 29.26 | 29.94 | 41.94 | 41.84 | 50.37 | 50.73 | |
| EmoVITCategory=Emotion-Oriented MLLM2025.12 | 78.5 | 78.24 | 28.81 | 28.8 | 41.61 | 38.09 | 49.64 | 48.38 | |
| Qwen2.5-VLCategory=Open-Source MLLM2025.12 | 77.8 | 77.76 | 28.38 | 31.35 | 36.97 | 37.43 | 47.72 | 48.85 | |
| InternVL3.5Category=Open-Source MLLM2025.12 | 76.8 | 76.66 | 27.95 | 28.63 | 40.82 | 41.8 | 48.52 | 49.03 | |
| GPT-5Category=Proprietary MLLM2025.12 | 64.2 | 63.02 | 41.92 | 43.48 | 45.53 | 45.33 | 50.55 | 50.61 | |
| Emotion-LLaMACategory=Emotion-Oriented MLLM2025.12 | 40.5 | 46.41 | 10.48 | 7.78 | 22.95 | 16.46 | 24.64 | 23.55 |