Back to Arena

AppAgent

by Tencent / mnotgod96

System Card

OrganizationTencent / mnotgod96
Released2023-12
Architectureknowledge-base / UI documentation knowledge base from exploration
DetailsMultimodal LLM agent that explores mobile apps (or learns from demos), generating UI documentation stored as a knowledge base used for later task execution via touchscreen actions. Works with GPT-4V / Qwen-VL.
Parameters—
Domainagent-memoryknowledge-graph
Open SourceYes
PaperView Paper
CodeRepository
mobile-agentui-docgpt-4vexploration

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
1/2
LoCoMo
73.137p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
73.434p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
7579p
HotpotQA
76.480p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
7594p
Sources:AppAgent paper (arXiv:2312.13771); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)AppAgent paper (arXiv:2312.13771); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)AppAgent paper (arXiv:2312.13771); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)AppAgent paper (arXiv:2312.13771); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)AppAgent paper (arXiv:2312.13771); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)AppAgent paper (arXiv:2312.13771); evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)