Conversational Memory Question Answering on LoCoMo
91.08Single-Hop J-ScoreEverMemOS
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EverMemOSTokens=2.5k, Evaluation Protocol=3-LLM-judge ensemble2026.05 | 91.08 | 86.17 | 81.93 | 66.67 | 86.76 | — | |
| COGNIFOLDTokens=4.2k2026.05 | 90.49 | 67.38 | 78.5 | 50 | 81.23 | 35.71 | |
| MemOSTokens=2.6k2026.05 | 81.09 | 67.49 | 75.18 | 55.9 | 75.8 | 45.27 | |
| ENGRAMTokens=0.9k2026.05 | 79.9 | 79.79 | 70.79 | 72.92 | 77.55 | 21.08 | |
| ZepTokens=2.7k2026.05 | 79.79 | 74.11 | 67.71 | 66.04 | 75.14 | 41.23 | |
| MemobaseTokens=2.1k2026.05 | 73.12 | 64.65 | 81.2 | 53.12 | 72.01 | 50.18 | |
| MIRIX2026.05 | 68.22 | 54.26 | 68.54 | 46.88 | 64.33 | 28.1 | |
| SupermemoryTokens=0.5k2026.05 | 67.3 | 51.12 | 31.77 | 42.67 | 55.34 | 34.87 | |
| Mem0Tokens=1.2k2026.05 | 67.13 | 51.15 | 55.51 | 72.93 | 66.88 | 43.46 | |
| MemUTokens=0.6k2026.05 | 66.34 | 63.12 | 27.1 | 50 | 56.55 | 35.15 |