Image-based Question Answering on WebQA
53.9AccuracyVimRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| VimRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 53.9 | |
| Vanilla RAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 48.1 | |
| VimRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 47.6 | |
| MemAgentBackbone=Qwen3-VL-8B-Instruct2026.02 | 47.1 | |
| MemAgentBackbone=Qwen3-VL-4B-Instruct2026.02 | 46.5 | |
| UniversalRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 45.8 | |
| Vanilla RAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 45.3 | |
| Mem1Backbone=Qwen3-VL-8B-Instruct2026.02 | 44.5 | |
| Mem1Backbone=Qwen3-VL-4B-Instruct2026.02 | 44.1 | |
| VideoRAGBackbone=Qwen3-VL-8B-Instruct2026.02 | 42.1 | |
| VideoRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 41.8 | |
| UniversalRAGBackbone=Qwen3-VL-4B-Instruct2026.02 | 41.2 | |
| ReActBackbone=Qwen3-VL-8B-Instruct2026.02 | 40 | |
| ReActBackbone=Qwen3-VL-4B-Instruct2026.02 | 39.3 |