Long-form audio-video understanding on LVOmniBench up to 90min
41.3Performance Score (Medium Tier)AVOC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AVOCLLM Backbone Scale=7-8B, Global token budget (K)=10240, Modality token ratio (Kvideo:Kaudio)=2:12026.06 | 41.3 | 35.5 | 40.3 | |
| MiniCPM-o 4.5LLM Backbone Scale=7-8B2026.06 | 34.1 | 25.1 | 34.8 | |
| MiniCPM-o 2.6LLM Backbone Scale=7-8B2026.06 | 33.8 | 26.6 | 32.5 | |
| OmniZipLLM Backbone Scale=7-8B2026.06 | 31.8 | 30.1 | 32.5 | |
| video-SALMONN 2+ 7BLLM Backbone Scale=7-8B2026.06 | 30.2 | 26.7 | 32.7 | |
| Qwen2.5-Omni-7BLLM Backbone Scale=7-8B2026.06 | 29.9 | 28.3 | 32 | |
| VideoLLaMA2-7BLLM Backbone Scale=7-8B2026.06 | 26.8 | 28.2 | 27.2 |