Question Answering on OBQA (Acc, Gap)
90.1AccuracySALAD-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SALAD-7BInput Modality=Text, Parameters=7B2025.10 | 90.1 | -1.1 | |
| Qwen2.5-Omni-7BInput Modality=Text, Parameters=7B2025.10 | 87.3 | 1.7 | |
| SALAD-3BInput Modality=Text, Parameters=3B2025.10 | 83.7 | -1.9 | |
| DiVA-Llama3.1-8BInput Modality=Text, Parameters=8B2025.10 | 82 | 1.3 | |
| Qwen2-Audio-7BInput Modality=Text, Parameters=7B2025.10 | 73.4 | 3.3 | |
| GLM-4-Voice-9BInput Modality=Text, Parameters=9B2025.10 | 69.9 | 17.8 | |
| MSRSBackbone=Mistral-7B-v0.32025.08 | 62.2 | — | |
| MSRSBackbone=Qwen2-7B-Instruct2025.08 | 61.6 | — | |
| VanillaBackbone=Qwen2-7B-Instruct2025.08 | 60.6 | — | |
| VanillaBackbone=Mistral-7B-v0.32025.08 | 60.2 | — | |
| MSRSBackbone=Llama-3-8B-Instruct2025.08 | 56.8 | — | |
| VanillaBackbone=Llama-3-8B-Instruct2025.08 | 55.6 | — |