List your company

Long-horizon Question Answering on LongMemEval-RR

46.33F1 ScoreMemAgent-7B
1.786813.350924.91536.4791Jun 4, 2026

Evaluation results

MethodLinks
MemAgent-7BQuery visible at write time?=Yes, Full raw-log access?=No, Source-evidence budget=native2026.0646.33——
TierMem with full raw logsQuery visible at write time?=No, Full raw-log access?=Yes, Source-evidence budget=full log2026.0645.4110061.52
Oracle retentionQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0644.9999.9883.67
Full raw-log RAGQuery visible at write time?=No, Full raw-log access?=Yes, Source-evidence budget=full log2026.0636.4510060.82
EMBER-14B writer (Qwen2.5-14B)Query visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0630.1732.1531.12
EMBER-7B writer (Qwen2.5-7B)Query visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0627.6829.6629.15
TierMem with budgeted raw fallbackQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0617.6512.469.97
Source-snippet heuristicQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0617.6326.5717.54
Memory-R1 (adapted)Query visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0615.6517.1414.64
Hybrid salienceQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0615.3821.417.63
MemAgent-7B (adapted)Query visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0613.11——
Generated-query indexingQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.0612.0520.2114.23
Summary-onlyQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=B-equiv.2026.0611.56——
Recency-BQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.068.2715.219.43
Reservoir-BQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.064.987.536.15
Random-BQuery visible at write time?=No, Full raw-log access?=No, Source-evidence budget=81922026.063.56.124.25