Evidence Retrieval on WalkieKnowledgeT+
66.83Retrieval Accuracy@1VL-MemKnG
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VL-MemKnG2026.06 | 66.83 | 72.3 | 75.35 | 40.55 | 57.67 | 60.82 | 66.83 | 36.9 | 24.49 | 66.8 | 69.4 | 70.1 | |
| Gemini 2.5Variant=Pro2026.06 | 66.03 | 85.06 | 86.35 | 40.5 | 59.01 | 60.68 | 66.02 | 35.29 | 22.3 | 66 | 74.9 | 75.2 | |
| Qwen 3.5+2026.06 | 63.37 | 78.59 | 81.67 | 39.69 | 60.93 | 65.05 | 63.37 | 38.07 | 25.36 | 63.4 | 70.5 | 71.2 | |
| Gemini 2.5Variant=Flash2026.06 | 59.01 | 77.6 | 78.78 | 37.22 | 53.21 | 54.53 | 59.01 | 31.42 | 19.48 | 59 | 67.3 | 67.6 | |
| VL-KnGVariant=GER-G2026.06 | 58.33 | 70.9 | 73.01 | 34.5 | 58.3 | 61.6 | 58.34 | 38.66 | 25.11 | 58.3 | 64 | 64.5 | |
| VL-KnGVariant=GER-L2026.06 | 55 | 70.9 | 74.94 | 32.22 | 57.75 | 62.5 | 55 | 38.23 | 25.25 | 55 | 62.4 | 63.4 |