Long-context memory-based conversation reasoning on BEAM 1M (test)
52.57Pass@0.5SelfMem
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SelfMemScale=1M2026.07 | 52.57 | 0.454 | 2.004 | 31.1 | 4,040 | 0 | |
| RAGScale=1M2026.07 | 38.3 | 0.32 | 1.843 | 0.4 | 3,468 | 74,153 | |
| Mem0Scale=1M2026.07 | 34.9 | 0.292 | 18.83 | 35.9 | 11,356 | 25,779 | |
| MemoryBankScale=1M2026.07 | 31.4 | 0.266 | 1.934 | 0.4 | 3,499 | 76,903 | |
| A-MemScale=1M2026.07 | 31.1 | 0.265 | 8.48 | 0.7 | 3,588 | 1,421 | |
| ReadAgentScale=1M2026.07 | 30.7 | 0.268 | 12.232 | 0.6 | 3,618 | 1,061 | |
| LoCoMoScale=1M2026.07 | 29.3 | 0.261 | 12.275 | 0.9 | 3,662 | 1,421 | |
| CompressionScale=1M2026.07 | 25.7 | 0.21 | 15.033 | 57.3 | 6,376 | 0 | |
| MemGPTScale=1M2026.07 | 25.7 | 0.221 | 12.264 | 0.8 | 3,609 | 989 |