Textual Emotion Analysis on DFEC internally curated 2025 (test)
9.01Test Score (2000 samples)FaVChat (Qwen3)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FaVChat (Qwen3)LLM Size=8B, Fine-tuned on FaVChat-170K=true2025.03 | 9.01 | 4.87 | 4.82 | 4.74 | 4.81 | 44.7 | 46.9 | 52.3 | |
| FaVChat (Qwen2.5)LLM Size=7B, Fine-tuned on FaVChat-170K=true2025.03 | 8.89 | 4.84 | 4.79 | 4.69 | 4.78 | 44.3 | 45.4 | 48.7 | |
| Qwen3-VL-8B-FaceLLM Size=8B, Fine-tuned on FaVChat-170K=true2025.03 | 8.87 | 4.58 | 4.46 | 4.47 | 4.51 | 29.7 | 31.2 | 42.9 | |
| Qwen2.5-VL-7B-FaceLLM Size=7B, Fine-tuned on FaVChat-170K=true2025.03 | 8.64 | 4.51 | 4.41 | 4.45 | 4.22 | 27.4 | 26.1 | 46.3 | |
| Qwen2.5-VL-72BLLM Size=72B2025.03 | 8.47 | 4.61 | 4.47 | 4.59 | 4.78 | 28.1 | 31.7 | 45.3 | |
| Qwen3-VL-30B-A3BLLM Size=30B2025.03 | 8.44 | 4.63 | 4.51 | 4.57 | 4.81 | 30.4 | 33.1 | 46.1 | |
| Qwen3-VL-8BLLM Size=8B2025.03 | 7.99 | 4.49 | 4.37 | 4.41 | 4.33 | 26.2 | 25.7 | 42.1 | |
| Qwen2.5-VL-7BLLM Size=7B2025.03 | 7.84 | 4.49 | 4.36 | 4.43 | 4.21 | 26.1 | 25.4 | 41.7 | |
| Claude4-Sonnet2025.03 | 7.84 | 4.55 | 4.64 | 4.77 | 4.84 | 31.2 | 34.2 | 47.2 | |
| GPT-4o2025.03 | 7.67 | 4.22 | 3.97 | 4.48 | 3.9 | 26.4 | 21.3 | 43.2 | |
| Omni-EmotionLLM Size=7B, Visual and speech modalities training=true2025.03 | 7.53 | 4.57 | 4.38 | 4.73 | 4.24 | 40.7 | 46.1 | 50.4 | |
| HumanOmniLLM Size=7B, Visual and speech modalities training=true2025.03 | 7.51 | 4.58 | 4.41 | 4.7 | 4.41 | 41.2 | 46.8 | 52.3 | |
| PLLaVALLM Size=7B2025.03 | 7.47 | 4.21 | 4.15 | 4.37 | 4.08 | 26.8 | 25 | 39.3 | |
| FaceTrack-MMLLM Size=7B2025.03 | 7.47 | 4.42 | 4.3 | 4.6 | 4.26 | 41.8 | 47.3 | 48.3 | |
| Emotion-LLaMALLM Size=7B, Visual and speech modalities training=true2025.03 | 7.42 | 4.17 | 4.33 | 4.54 | 4.19 | 29.4 | 41.3 | 37.2 | |
| VideoLLaMA2LLM Size=7B2025.03 | 7.41 | 4.17 | 4.02 | 4.47 | 3.93 | 25.3 | 26.6 | 34.4 | |
| ShareGPT4VideoLLM Size=7B2025.03 | 7.33 | 4.24 | 4.13 | 4.35 | 4.09 | 19.2 | 20.5 | 39.4 |