Multi-task Language Understanding on MMSU
71.6AccuracySALAD-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SALAD-7BInput Modality=Text, Parameters=7B2025.10 | 71.6 | -0.8 | |
| ASR + Qwen2.5-7BSystem Category=Cascaded Toplines, ASR=Whisper-v3-Large, Backbone LLM=Qwen2.5-7B2026.05 | 67.9 | 2.9 | |
| TextPro-SLM-7BSystem Category=Modality Gap-optimized Systems2026.05 | 66.7 | 4.2 | |
| Qwen2.5-Omni-7BInput Modality=Text, Parameters=7B2025.10 | 66.3 | 4.5 | |
| TextPro-SLM-7B 5:1System Category=Modality Gap-optimized Systems, Sampling Ratio=5:12026.05 | 66.2 | 4.7 | |
| WhisperPro + Qwen2.5-7BSystem Category=Ablation Studies, Backbone LLM=Qwen2.5-7B2026.05 | 62.6 | 8.3 | |
| SALAD-3BInput Modality=Text, Parameters=3B2025.10 | 62.5 | -0.6 | |
| Kimi-Audio-7BSystem Category=End-to-end Systems, Backbone LLM=Kimi-Audio-7B2026.05 | 61.6 | 9.2 | |
| Qwen2.5-Omni-7BSystem Category=End-to-end Systems, Backbone LLM=Qwen2.5-Omni-7B2026.05 | 61.3 | 9.6 | |
| DiVA-Llama3.1-8BInput Modality=Text, Parameters=8B2025.10 | 60.9 | 1 | |
| ASR + Qwen2.5-3BSystem Category=Cascaded Toplines, ASR=Whisper-v3-Large, Backbone LLM=Qwen2.5-3B2026.05 | 58.9 | 2.7 | |
| TextPro-SLM-3BSystem Category=Modality Gap-optimized Systems2026.05 | 57.6 | 4 | |
| SALAD-7BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage II2026.05 | 57.5 | 13.3 | |
| SALAD-7BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage I2026.05 | 55.3 | 15.5 | |
| SALAD-3BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage II2026.05 | 52.5 | 9.4 | |
| GLM-4-Voice-9BInput Modality=Text, Parameters=9B2025.10 | 48.4 | 17.8 | |
| SALAD-3BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage I2026.05 | 47.3 | 14.6 | |
| Qwen2-Audio-7BInput Modality=Text, Parameters=7B2025.10 | 46.9 | 1.3 | |
| TextPro-SLM-7B w/o KDSystem Category=Ablation Studies, Ablation Setting=w/o KD2026.05 | 46.7 | 24.2 | |
| GLM-4-Voice-9BSystem Category=End-to-end Systems, Backbone LLM=GLM-4-Voice-9B2026.05 | 39.2 | 28.3 | |
| Qwen2-Audio-7BSystem Category=End-to-end Systems, Backbone LLM=Qwen2-Audio-7B2026.05 | 29.5 | 18.7 | |
| DiVA-Llama3.1-8BSystem Category=End-to-end Systems, Backbone LLM=Llama3.1-8B2026.05 | 28.8 | 34.1 | |
| Random2026.05 | 25 | — |