Video Corpus Understanding on SyntheticQA
54.5AccuracyVimRAG
Evaluation results
| Method | Links | |
|---|---|---|
| VimRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 54.5 | |
| VimRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 51 | |
| VideoRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 49.5 | |
| VideoRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 45.8 | |
| Mem1Backbone=Qwen3-VL-8B-Instruct2026.02 | 43.4 | |
| Mem1Backbone=Qwen3-VL-4B-Instruct2026.02 | 42.9 | |
| MemAgentBackbone=Qwen3-VL-8B-Instruct2026.02 | 37.5 | |
| Vanilla RAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 37 | |
| ReActBackbone=Qwen3-VL-8B-Instruct2026.02 | 35 | |
| MemAgentBackbone=Qwen3-VL-4B-Instruct2026.02 | 34.3 | |
| Vanilla RAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 32.6 | |
| ReActBackbone=Qwen3-VL-4B-Instruct2026.02 | 29.9 | |
| UniversalRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 25 | |
| UniversalRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 23.9 |