SoMem: agents organize their own linked-document memory, outscoring baselines on three benchmarks
Overview
Researchers from City University of Hong Kong and Tencent Jarvis Lab present SoMem, a framework in which LLM agents store memory as ordinary linked documents instead of fixed schemas.
The agent decides what to summarize and links each summary to its original interaction. SoMem beat the strongest baseline by 2.66 points on LoCoMo, 3.10 on BEAM-100K and 2.00 on LongMemEval-S. Elvis Saravia, who reported the study, says he uses a similar linked-document approach with his own agents.
Written by AI from the articles below · updated Nov 12, 10:05 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
elvis@omarsar0XResearchers test whether LLM agents can organize their own memoryAISoMem, a framework from City University of Hong Kong and Tencent Jarvis Lab, lets agents store memory as ordinary linked documents rather than fixed schemas. The agent decides what to summarize and links each summary to its original interaction. SoMem beat the strongest baseline by 2.66 points on LoCoMo, 3.10 on BEAM-100K, and 2.00 on LongMemEval-S. The author, Elvis Saravia, says he uses a similar linked-document approach with his own agents.

Heat Trend
- Current comparable heat
- 6
- Peak comparable heat
- 10Nov 12, 11:00 AM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.