Back to Arena

AgentVerse

by OpenBMB (Tsinghua)

System Card

OrganizationOpenBMB (Tsinghua)
Released2023-05
Architectureagentic-workflow / Multi-agent task-solving + simulation with shared memory
DetailsDual frameworks for task-solving (cooperative agents on problems) and simulation (open environments for emergent behavior). Agents share conversation memory in simulation contexts; ICLR 2024.
Parameters—
Domainagent-memoryepisodic-session
Open SourceYes
PaperView Paper
CodeRepository
multi-agentsimulationiclr2024openbmb

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
75.148p
MemoryBank
76.756p
Cross-Session Memory
1/1
LongMemEval
80.684p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
72.860p
HotpotQA
77.688p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:AgentVerse paper (arXiv:2308.10848); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)AgentVerse paper (arXiv:2308.10848); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)AgentVerse paper (arXiv:2308.10848); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)AgentVerse paper (arXiv:2308.10848); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)AgentVerse paper (arXiv:2308.10848); evaluated on MemoryBank: Enhancing LLMs with Long-Term Memory (Sun Yat-sen University, 2305)AgentVerse paper (arXiv:2308.10848); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)