Paralinguistic Understanding on Paralinguistic Understanding Tasks (test)
61.1Emotion AccuracyKimi-Audio-7B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Kimi-Audio-7BBackbone=7B2026.05 | 61.1 | 75.9 | 60.5 | 31.2 | 57.2 | |
| TextPro-SLM-7BBackbone=7B2026.05 | 60.5 | 88.6 | 64.9 | 45.1 | 64.8 | |
| TextPro-SLM-3BBackbone=3B2026.05 | 58.8 | 80.2 | 63.7 | 45 | 61.9 | |
| TextPro-SLM-7B 5:1Backbone=7B, Prosody Embedding Ratio=5:12026.05 | 57.7 | 88.8 | 66.7 | 50.2 | 65.8 | |
| TextPro-SLM-7B w/o ReconBackbone=7B, Ablation variant=Without reconstruction objective2026.05 | 56.7 | 79.9 | 64.2 | 44.4 | 61.3 | |
| Qwen2.5-Omni-7BBackbone=7B2026.05 | 54.8 | 89.8 | 44.8 | 45.3 | 58.7 | |
| TextPro-SLM-7B w/o TrainBackbone=7B, Ablation variant=Without paralinguistic training2026.05 | 51.8 | 28.5 | 43.6 | 16.1 | 35 | |
| Qwen2-Audio-7BBackbone=7B2026.05 | 48.2 | 64.7 | 23.1 | 29 | 41.3 | |
| DiVA2026.05 | 30.7 | 46.8 | 33.3 | 32.5 | 35.8 | |
| GLM-4-Voice-9BBackbone=9B2026.05 | 23 | 23.9 | 18.7 | 32.4 | 24.5 |