Visual Question Answering on Visual Memory Benchmark
100Target Person AccuracyFull context
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Full contextTokens=325K2026.05 | 100 | 94.6 | 91.4 | 98 | 95.1 | |
| OracleTokens=19002026.05 | 100 | 99.1 | 97.9 | 98.5 | 98.6 | |
| VISUALMEMTokens=19802026.05 | 95 | 91.1 | 77.9 | 83.4 | 84.1 | |
| MemOSTokens=11872026.05 | 45 | 59.9 | 52.1 | 64.8 | 56 | |
| Mem0Tokens=5002026.05 | 38 | 33.9 | 40.8 | 57.9 | 45 | |
| LightMemTokens=5002026.05 | 30 | 40.2 | 45.4 | 58.8 | 46.1 | |
| HippoRAG2Tokens=10002026.05 | 25 | 26.5 | 25.4 | 32.7 | 27.6 | |
| Self-RAGTokens=20002026.05 | 21 | 35 | 17.1 | 22.1 | 22.1 | |
| SimpleMemTokens=5002026.05 | 3 | 40.2 | 42.9 | 43.2 | 36.8 |