Question Answering on ARC Challenge (Accuracy (%) and Modality Gap)
89.4AccuracyTextPro-SLM-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TextPro-SLM-7BSystem Category=Modality Gap-optimized Systems2026.05 | 89.4 | -0.2 | |
| TextPro-SLM-7B 5:1System Category=Modality Gap-optimized Systems, Sampling Ratio=5:12026.05 | 88.8 | 0.4 | |
| Kimi-Audio-7BSystem Category=End-to-end Systems, Backbone LLM=Kimi-Audio-7B2026.05 | 88.2 | 1 | |
| WhisperPro + Qwen2.5-7BSystem Category=Ablation Studies, Backbone LLM=Qwen2.5-7B2026.05 | 87.8 | 1.4 | |
| Qwen2.5-Omni-7BSystem Category=End-to-end Systems, Backbone LLM=Qwen2.5-Omni-7B2026.05 | 86.1 | 3 | |
| SALAD-7BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage II2026.05 | 84 | 4.4 | |
| ASR + Qwen2.5-7BSystem Category=Cascaded Toplines, ASR=Whisper-v3-Large, Backbone LLM=Qwen2.5-7B2026.05 | 83.3 | 5.8 | |
| SALAD-7BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage I2026.05 | 82.3 | 6.1 | |
| TextPro-SLM-3BSystem Category=Modality Gap-optimized Systems2026.05 | 80.7 | 3.1 | |
| ASR + Qwen2.5-3BSystem Category=Cascaded Toplines, ASR=Whisper-v3-Large, Backbone LLM=Qwen2.5-3B2026.05 | 80.1 | 3.7 | |
| SALAD-3BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage II2026.05 | 79.9 | 1.9 | |
| TextPro-SLM-7B w/o KDSystem Category=Ablation Studies, Ablation Setting=w/o KD2026.05 | 75.8 | 13.4 | |
| SALAD-3BSystem Category=Modality Gap-optimized Systems, Training Stage=Stage I2026.05 | 75.6 | 6.2 | |
| GLM-4-Voice-9BSystem Category=End-to-end Systems, Backbone LLM=GLM-4-Voice-9B2026.05 | 59.5 | 33.8 | |
| Qwen2-Audio-7BSystem Category=End-to-end Systems, Backbone LLM=Qwen2-Audio-7B2026.05 | 43.5 | 28.5 | |
| DiVA-Llama3.1-8BSystem Category=End-to-end Systems, Backbone LLM=Llama3.1-8B2026.05 | 33.7 | 48.9 | |
| Random2026.05 | 25 | — |