Back to Arena

Synapse

by Nanyang Technological University (Zheng et al.)

System Card

OrganizationNanyang Technological University (Zheng et al.)
Released2023-06
Architectureagentic-workflow / Trajectory-as-exemplar prompting with exemplar memory
DetailsThree components: state abstraction (compresses raw HTML into concise task observations), trajectory-as-exemplar prompting (uses full trajectory sequences as demonstrations), and exemplar memory retrieved via similarity search.
Parameters—
Domainagent-memoryepisodic-session
Open SourceYes
PaperView Paper
WebsiteVisit
CodeRepository
web-agentstrajectory-memoryminiwobmind2webcomputer-control

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Data Transparency:6 estimated
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
77.571pEstimated
MemoryBank
73.432pEstimated
Cross-Session Memory
1/1
LongMemEval
79.373pEstimated
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
70.338pEstimated
HotpotQA
77.285pEstimated
Agent Task Memory
1/1
AgentBench-Mem
7226pEstimated
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:Arena estimate — derived from capability profile, not independently verified