Back to Arena

GPTeam

by 101dotxyz

System Card

Organization101dotxyz
Released2023-06
Architectureepisodic-buffer / Generative-agents clone with memory + reflection
DetailsOpen-source simulation of GPT-4 agents collaborating across a virtual map. Agents have individual memory, reflection, and inter-agent messaging; directly inspired by Stanford Generative Agents.
Parameters—
Domainagent-memoryepisodic-session
Open SourceYes
CodeRepository
generative-agentssimulationreflectiontypescript

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
75.451p
MemoryBank
78.369p
Cross-Session Memory
1/1
LongMemEval
77.964p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
69.735p
HotpotQA
68.940p
Agent Task Memory
1/1
AgentBench-Mem
71.424p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:GPTeam (101dotxyz/GPTeam); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)GPTeam (101dotxyz/GPTeam); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)GPTeam (101dotxyz/GPTeam); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)GPTeam (101dotxyz/GPTeam); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)GPTeam (101dotxyz/GPTeam); evaluated on MemoryBank: Enhancing LLMs with Long-Term Memory (Sun Yat-sen University, 2305)GPTeam (101dotxyz/GPTeam); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)