Overall Performance on LoCoMo
48.7BLEUFull Context
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Full Context2026.02 | 48.7 | 56.5 | 82.5 | |
| MEMORARetriever Type=Policy Retriever2026.02 | 46.6 | 55.3 | 86.3 | |
| MEMORARetriever Type=Semantic Retriever2026.02 | 46.4 | 55.2 | 84.9 | |
| Nemori*2026.02 | 45.6 | 53.4 | 79.4 | |
| R2-MemModel=Qwen2.5-7B2026.05 | 44.91 | 51.35 | — | |
| R2-MemModel=Llama3.1-8B2026.05 | 40.66 | 47.13 | — | |
| LangMem*2026.02 | 40 | 47.6 | 73.4 | |
| GAMModel=Qwen2.5-7B2026.05 | 39.42 | 45.77 | — | |
| RAG2026.02 | 38.9 | 45.5 | 63.3 | |
| GAMModel=Llama3.1-8B2026.05 | 37.48 | 44.15 | — | |
| Mem02026.02 | 34.6 | 41.1 | 65.3 | |
| Zep*2026.02 | 30.9 | 36.9 | 61.6 | |
| Memory R1Model=Llama3.1-8B2026.05 | 29.51 | 37.13 | — | |
| Memory R1Model=Qwen2.5-7B2026.05 | 27.49 | 36.07 | — | |
| LIGHTMEMModel=Llama3.1-8B2026.05 | 25.51 | 30.01 | — | |
| MemoryOSModel=Llama3.1-8B2026.05 | 24.54 | 33.05 | — | |
| LIGHTMEMModel=Qwen2.5-7B2026.05 | 23.67 | 29.9 | — | |
| MemoryOSModel=Qwen2.5-7B2026.05 | 23.59 | 31.45 | — | |
| Mem0Model=Qwen2.5-7B2026.05 | 20.76 | 27.38 | — | |
| Mem0Model=Llama3.1-8B2026.05 | 20.55 | 27.83 | — | |
| A-MEMModel=Llama3.1-8B2026.05 | 15.23 | 21.09 | — | |
| A-MEMModel=Qwen2.5-7B2026.05 | 13.91 | 19.63 | — | |
| RAGModel=Llama3.1-8B2026.05 | 9.44 | 11.81 | — | |
| RAGModel=Qwen2.5-7B2026.05 | 8.17 | 10.1 | — |