Long-Context Memory Retrieval and Reasoning on LoCoMo 10K context input token limit
78.1Single-hop L-J ScoreNo limit
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| No limitPrimary Agent Backbone=Claude Sonnet 4.5, Context Token Limit=No limit2026.06 | 78.1 | 66.4 | 1.5 | 55.9 | |
| HORMAPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=External Dynamic Memory, Context Token Limit=10K2026.06 | 70.5 | 50.5 | 13.4 | 51.6 | |
| Mem0Primary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=External Dynamic Memory, Context Token Limit=10K2026.06 | 64.7 | 28 | 11.2 | 43.4 | |
| Embedding RetrievalPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=External Dynamic Memory, Context Token Limit=10K2026.06 | 57.2 | 33.6 | 0.7 | 37.8 | |
| TruncationPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=Internal Static Memory, Context Token Limit=10K2026.06 | 47.8 | 30.8 | 0.7 | 32.2 | |
| EmbeddingPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=Internal Static Memory, Context Token Limit=10K2026.06 | 47.5 | 45.8 | 0 | 34.8 | |
| AconPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=Internal Dynamic Memory, Context Token Limit=10K2026.06 | 42.8 | 26.2 | 16.4 | 32.6 | |
| Slide WindowPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=Internal Static Memory, Context Token Limit=10K2026.06 | 39.9 | 47.7 | 0.7 | 31.4 | |
| A-MEMPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=External Dynamic Memory, Context Token Limit=10K2026.06 | 39.6 | 35.5 | 7.5 | 30.4 | |
| ReSumPrimary Agent Backbone=Claude Sonnet 4.5, Memory Architecture Category=Internal Dynamic Memory, Context Token Limit=10K2026.06 | 21.9 | 32.7 | 0.7 | 18.7 |