Back to Arena

BabyAGI

by Yohei Nakajima

System Card

OrganizationYohei Nakajima
Released2023-01
Architectureagentic-workflow / Self-building "functionz" graph
DetailsNew BabyAGI is a `functionz` framework: a graph-based store of agent functions with dependency tracking, import management, and AI-powered code generation that lets the agent author new functions.
Parameters—
Domainagent-memorylifelong-learning
Open SourceYes
CodeRepository
functionzself-buildinggraphcanonical

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
75.655p
MemoryBank
74.741p
Cross-Session Memory
1/1
LongMemEval
80.986p
Multi-Hop QA
2/3
BABILong
73.129p
MultiHop-RAG
no data
HotpotQA
72.657p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:BabyAGI (yoheinakajima/babyagi); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)BabyAGI (yoheinakajima/babyagi); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)BabyAGI (yoheinakajima/babyagi); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)BabyAGI (yoheinakajima/babyagi); evaluated on BABILong: Testing the Limits of LLMs with Long-Context Reasoning-in-a-Haystack (AIRI, 2406)BabyAGI (yoheinakajima/babyagi); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)BabyAGI (yoheinakajima/babyagi); evaluated on MemoryBank: Enhancing LLMs with Long-Term Memory (Sun Yat-sen University, 2305)