Multimodal Question Answering on MAviS-Bench
34.17ROUGE-1MAviS-Chat
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MAviS-ChatEvaluation Setting=Fine-tuned, Vision Encoder=SigLip-400M, Audio Encoder=Whisper-medium-300M, Language Model=Qwen2.5-7B, Parameter Count=~8B2026.03 | 34.17 | 29.31 | 87.42 | 54.76 | 61.1 | 59.92 | |
| Phi-4-MM-InstructEvaluation Setting=Zero-shot or few-shot2026.03 | 32.43 | 30.18 | 88.68 | 54.42 | 48.04 | 57.7 | |
| GPT-4oEvaluation Setting=Zero-shot or few-shot2026.03 | 30.55 | 34.08 | 87.92 | 54.03 | 59.88 | 71.18 | |
| GPT-4o-miniEvaluation Setting=Zero-shot or few-shot2026.03 | 24.28 | 29.72 | 86.52 | 52.58 | 60.04 | 69.16 | |
| MiniCPM-o-2.6Evaluation Setting=Zero-shot or few-shot2026.03 | 19.77 | 27.14 | 85.33 | 52.1 | 55.46 | 52.42 | |
| Gemini 1.5Evaluation Setting=Zero-shot or few-shot2026.03 | 18.95 | 23.31 | 84.42 | 50.63 | 45.16 | 47.36 |