Long Multi-session Question Answering on LongMemEval 1.0 (test)
34.6KU F1HORMA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HORMABackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=External Dynamic Memory2026.06 | 34.6 | 15.8 | 10.7 | 56.9 | 17.3 | 30.2 | |
| AconBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=Internal Dynamic Memory2026.06 | 24.5 | 13.3 | 15.2 | 32.7 | 12.6 | 25.3 | |
| ReSumBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=Internal Dynamic Memory2026.06 | 23.2 | 12.7 | 12.5 | 30 | 9.1 | 16.9 | |
| A-MEMBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=External Dynamic Memory2026.06 | 21.7 | 15.5 | 18.8 | 30 | 10.6 | 18.6 | |
| TruncationBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=Internal Static Memory2026.06 | 21.4 | 10.6 | 13.5 | 13.9 | 6.9 | 12.4 | |
| Mem0Backbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=External Dynamic Memory2026.06 | 16.3 | 14.6 | 14.2 | 35.3 | 9 | 15.7 | |
| Embedding RetrievalBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=External Dynamic Memory2026.06 | 13.7 | 6.6 | 14.2 | 17.7 | 7.3 | 11.2 | |
| No limitBackbone=Claude Sonnet 4.5, Context Limit=Unconstrained2026.06 | 9.2 | 6.2 | 12 | 4.5 | 6 | 6.9 | |
| EmbeddingBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=Internal Static Memory2026.06 | 8.8 | 7.7 | 11 | 7.9 | 7.8 | 8.2 | |
| Slide WindowBackbone=Claude Sonnet 4.5, Context Limit=50K tokens, Memory Architecture=Internal Static Memory2026.06 | 7.1 | 5.4 | 9.8 | 7.5 | 7.3 | 7.2 |