Long-context memory evaluation on MemBench
87.5RecallEVOLVEMEM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EVOLVEMEMBackbone=GPT-4o2026.05 | 87.5 | 66.7 | 50 | 67.9 | |
| EVOLVEMEMBackbone=GPT-5.12026.05 | 87.5 | 66.7 | 62.5 | 71.4 | |
| RecentMemBackbone=GPT-5.12026.05 | 75 | 50 | 62.5 | 60.7 | |
| MemGPTBackbone=GPT-5.12026.05 | 75 | 58.3 | 50 | 60.7 | |
| MemBankBackbone=GPT-5.12026.05 | 75 | 50 | 75 | 64.3 | |
| RecentMemBackbone=GPT-4o2026.05 | 62.5 | 50 | 62.5 | 57.1 | |
| MemGPTBackbone=GPT-4o2026.05 | 62.5 | 50 | 62.5 | 57.1 | |
| SCMemBackbone=GPT-4o2026.05 | 62.5 | 25 | 37.5 | 39.3 | |
| SCMemBackbone=GPT-5.12026.05 | 50 | 25 | 25 | 32.1 | |
| MemBankBackbone=GPT-4o2026.05 | 37.5 | 33.3 | 75 | 46.4 |