Back to Arena

CAMEL

by CAMEL-AI.org

System Card

OrganizationCAMEL-AI.org
Released2023-05
Architectureagentic-workflow / Stateful memory + storage + retrieval for agent societies
DetailsAgent + Society framework with core modules for Agents, Data Generation, Models, Tools, Memory, Storage, and graph-based retrieval. Emphasizes scaling laws for agent societies up to millions of agents.
Parameters—
Domainagent-memory
Open SourceYes
PaperView Paper
WebsiteVisit
CodeRepository
agent-societyscaling-lawsrole-playingapache-2

Capability Profile

Benchmark Scores

5 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
1/2
LoCoMo
77.268p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
79.475p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
71.144p
HotpotQA
75.674p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:CAMEL paper (arXiv:2303.17760); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)CAMEL paper (arXiv:2303.17760); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)CAMEL paper (arXiv:2303.17760); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)CAMEL paper (arXiv:2303.17760); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)CAMEL paper (arXiv:2303.17760); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)