Back to Arena

Memori

by GibsonAI

System Card

OrganizationGibsonAI
Released2025-03
Architecturehierarchical-summary / SQL-backed conversational memory
DetailsStores conversational memory as structured rows in a managed SQL database with periodic LLM-driven consolidation passes. Pluggable SQL backend (SQLite/Postgres) keeps writes auditable.
Parameters—
Domainagent-memorypersonalization
Open SourceYes
WebsiteVisit
CodeRepository
sql-backedconsolidationstructured

Capability Profile

Benchmark Scores

6 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
54.82p
MemoryBank
72.628p
Cross-Session Memory
1/1
LongMemEval
54.81p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
67.629p
HotpotQA
6530p
Agent Task Memory
1/1
AgentBench-Mem
56.91p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:Memori internal evalMemori (GibsonAI/memori); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)Memori (GibsonAI/memori); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)Memori (GibsonAI/memori); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Memori (GibsonAI/memori); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)