Long-horizon Agent Performance on CAUSAL-LOCOMO
84.6Task ScoreCMI
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CMIAvg. Memories=0.9432026.05 | 84.6 | 81.6 | 87.5 | 99 | 0 | |
| Reflection MemoryAvg. Memories=3.0002026.05 | 84.5 | 79.3 | 48.6 | 55.7 | 54 | |
| Vector MemoryAvg. Memories=3.0002026.05 | 83.9 | 78.2 | 50.1 | 56.6 | 60.9 | |
| Graph MemoryAvg. Memories=3.0002026.05 | 82.4 | 75.9 | 46.9 | 55 | 58.6 | |
| Summary MemoryAvg. Memories=5.6442026.05 | 72.3 | 58.6 | 30.8 | 0 | 62.1 | |
| Full HistoryAvg. Memories=5.6442026.05 | 51.5 | 21.8 | 30.8 | 0 | 62.1 | |
| No MemoryAvg. Memories=0.0002026.05 | 42.9 | 3.4 | 0 | 100 | 0 |