Agentic Memory Recall on LoCoMo
76.7Accuracy (%)Original
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OriginalModel=Qwen3-VL-235B-A22B2026.02 | 76.7 | 120 | |
| OriginalModel=Qwen2-VL-72B-Instruct2026.02 | 76 | 120 | |
| OriginalModel=Qwen2.5-VL-7B-Instruct2026.02 | 73 | 120 | |
| VOIModel=Qwen3-VL-235B-A22B2026.02 | 72.6 | 51.5 | |
| VOIModel=Qwen2-VL-72B-Instruct2026.02 | 68.1 | 40.8 | |
| VOIModel=Qwen2.5-VL-7B-Instruct2026.02 | 67.1 | 52.3 | |
| OriginalModel=LLaVA-1.5-7B-hf2026.02 | 64.7 | 120 | |
| VOIModel=LLaVA-1.5-7B-hf2026.02 | 61.5 | 53 | |
| OriginalModel=Llama-4-Maverick(17Bx12E)2026.02 | 60.8 | 120 | |
| VOIModel=Llama-4-Maverick(17Bx12E)2026.02 | 56.2 | 50.9 | |
| OriginalModel=Pixtral-12B-24092026.02 | 46.7 | 120 | |
| CaptionModel=LLaVA-1.5-7B-hf2026.02 | 43.7 | 10 | |
| VOIModel=Pixtral-12B-24092026.02 | 43.6 | 69.5 | |
| CaptionModel=Pixtral-12B-24092026.02 | 32.5 | 10 | |
| CaptionModel=Llama-4-Maverick(17Bx12E)2026.02 | 29.3 | 10 | |
| CaptionModel=Qwen3-VL-235B-A22B2026.02 | 27.4 | 10 | |
| CaptionModel=Qwen2.5-VL-7B-Instruct2026.02 | 26.8 | 10 | |
| CaptionModel=Qwen2-VL-72B-Instruct2026.02 | 26.4 | 10 |