Back to Arena

JARVIS-1

by CraftJarvis

System Card

OrganizationCraftJarvis
Released2023-10
Architectureepisodic-buffer / Multimodal memory for open-world planning
DetailsMinecraft agent combining a multimodal LM with embodied control and a multimodal memory store. Plans using both pre-trained knowledge and retrieved game-experience memories across 200+ tasks.
Parameters—
Domainagent-memorylifelong-learning
Open SourceYes
PaperView Paper
CodeRepository
minecraftmultimodal-memoryopen-worldpaper-repo

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
71.728p
MemoryBank
79.877p
Cross-Session Memory
1/1
LongMemEval
79.879p
Multi-Hop QA
2/3
BABILong
71.616p
MultiHop-RAG
no data
HotpotQA
70.648p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:JARVIS-1 paper (arXiv:2311.05997); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)JARVIS-1 paper (arXiv:2311.05997); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)JARVIS-1 paper (arXiv:2311.05997); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)JARVIS-1 paper (arXiv:2311.05997); evaluated on BABILong: Testing the Limits of LLMs with Long-Context Reasoning-in-a-Haystack (AIRI, 2406)JARVIS-1 paper (arXiv:2311.05997); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)JARVIS-1 paper (arXiv:2311.05997); evaluated on MemoryBank: Enhancing LLMs with Long-Term Memory (Sun Yat-sen University, 2305)