Back to Arena

Mem0

by Mem0

System Card

OrganizationMem0
Released2024-03
Architecturehybrid / Vector + LLM-extracted facts
DetailsTwo-phase pipeline: an LLM distills each turn into structured fact memories, which are stored in a vector index alongside raw episodes. Retrieval merges semantic similarity with rule-based filters and decay.
Parameters—
Domainagent-memorypersonalizationepisodic-session
Open SourceYes
PaperView Paper
WebsiteVisit
CodeRepository
llm-extractedvectorpersonalizationagentopen-source

Capability Profile

Benchmark Scores

3 of 14 benchmarks
Long-Context Retrieval
0/5
RULER
no data
NIAH
no data
LooGLE
no data
LongBench
no data
∞Bench
no data
Multi-Turn Recall
2/2
LoCoMo
68.522p
MemoryBank
7867p
Cross-Session Memory
1/1
LongMemEval
66.921p
Multi-Hop QA
0/3
BABILong
no data
MultiHop-RAG
no data
HotpotQA
no data
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:Mem0 technical reportMem0 internal evals