Emotion Recognition on MELD (held-out)
71.1F1 ScoreOMNISAPIENS-7B SFT
Evaluation results
| Method | Links | |
|---|---|---|
| OMNISAPIENS-7B SFTfine-tuning epochs=1, training mode=SFT2025.10 | 71.1 | |
| Qwen 2.5-Omni-7B SFTfine-tuning epochs=1, training mode=SFT2025.10 | 68.4 | |
| Qwen 2.5-Omni-7Bfine-tuning epochs=0, training mode=pre-trained2025.10 | 66.1 | |
| Gemma-3-4Bfine-tuning epochs=0, training mode=pre-trained2025.10 | 64.2 | |
| HumanOmniV2-7Bfine-tuning epochs=0, training mode=pre-trained2025.10 | 63.3 | |
| Qwen-2.5-VL-7Bfine-tuning epochs=0, training mode=pre-trained2025.10 | 57.1 | |
| OMNISAPIENS-7B RLzero-shot=true2025.10 | 54.9 | |
| Qwen 2.5-Omni-7Bzero-shot=true2025.10 | 40.3 |