MER and MSA Evaluation Suite
79.09Accuracy (MER2023)Nano-EmoX
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Nano-EmoXScale=2.2B, training_framework=P2E2026.03 | 79.09 | 77.94 | 56.55 | 60.12 | 76.82 | 79.81 | 86.25 | 84.76 | 64.75 | 74.01 | |
| AffectGPTScale=↑ 6.1B2026.03 | 78.54 | 78.8 | 55.65 | 60.54 | 81.3 | 80.9 | 88.49 | 86.18 | 62.52 | 74.77 | |
| Nano-EmoXScale=2.2B, training_approach=joint training2026.03 | 74.26 | 78.61 | 54.27 | 61.54 | 80.71 | 79.52 | 84.64 | 83.31 | 62.68 | 73.28 | |
| AffectGPT (s)Scale=↓ 0.1B, training_data=MER-caption+2026.03 | 73.45 | 74.71 | 47.69 | 53.14 | 75.51 | 71.3 | 82.5 | 84.1 | 62.43 | 69.43 | |
| AffectGPT (s)Scale=↓ 0.1B, training_framework=P2E2026.03 | 72.43 | 77.83 | 50.19 | 57.64 | 80.4 | 79.97 | 83.28 | 83.23 | 63.75 | 72.08 | |
| Qwen-2VL-7BScale=↑ 5.5B2026.03 | 59.81 | 69.14 | 48.05 | 50.53 | 74.1 | 58.35 | 78.65 | 77.43 | 55.61 | 63.52 | |
| Emotion-LLaMAScale=↑ 5.6B2026.03 | 59.38 | 73.62 | 46.76 | 55.47 | 66.13 | 67.66 | 78.32 | 77.23 | 52.97 | 64.17 | |
| R1-OmniScale=↓ 0.1B2026.03 | 58.3 | 69.41 | 40.87 | 50.18 | 55.56 | 48.62 | 74.71 | 76.67 | 51.84 | 58.46 | |
| SALMONNScale=↑ 11.7B2026.03 | 55.53 | 45.38 | 45.62 | 46.84 | 81 | 67.03 | 68.69 | 68.69 | 45 | 57.89 | |
| MiniCPM-V-2.6-8BScale=↑ 5.8B2026.03 | 46.67 | 45.31 | 40.27 | 36.31 | 74.96 | 57.44 | 74.85 | 75.04 | 50.04 | 55.65 | |
| MobileVLM V2-3BScale=↑ 0.8B2026.03 | 37.7 | 53.87 | 30.72 | 53 | 55.81 | 44.87 | 69.17 | 65.72 | 33.86 | 49.95 | |
| MobileVLM V2-1.7BScale=↓ 0.5B2026.03 | 36.65 | 47.03 | 33.37 | 49.24 | 41 | 56.49 | 51.46 | 51.94 | 36.96 | 44.9 |