Back to Arena

Kindroid

by Kindroid

System Card

OrganizationKindroid
Released2023-06
Architecturehierarchical-summary / Five-Layer Cascaded Memory
DetailsKindroid uses a five-layer configurable memory architecture combining user-provided facts, AI-generated inferences, behavioral pattern tracking, emotional preference modeling, and relationship dynamics — all updated via a Cascaded Memory system for context-aware retrieval. Supports rich backstory definition, AI-generated selfies, voice calls, and Second Life integration. Scored second-highest in independent memory recall tests (14/25 detailed facts retained vs. Nomi's 23/25). Founded by Stanford CS dropout Jerry Meng, launched June 2023.
Parameters—
Domainpersonalizationepisodic-sessionlifelong-learning
Open SourceNo
WebsiteVisit
companionfive-layer-memoryvoice-callscustomizableavatarsdeep-personalization

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Data Transparency:6 estimated
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
77.873pEstimated
MemoryBank
75.349pEstimated
Cross-Session Memory
1/1
LongMemEval
70.423pEstimated
Multi-Hop QA
1/3
BABILong
77.171pEstimated
MultiHop-RAG
no data
HotpotQA
no data
Agent Task Memory
1/1
AgentBench-Mem
71.524pEstimated
Personalization
1/1
PerLTQA
74.636pEstimated
Factuality / Grounding
0/1
RAGAS
no data
Sources:Arena estimate — derived from capability profile, not independently verified