Back to Arena

ChatDB

by Tsinghua University (Hu et al.)

System Card

OrganizationTsinghua University (Hu et al.)
Released2023-06
Architectureknowledge-base / SQL database as symbolic memory with chain-of-memory
DetailsUses SQL databases as symbolic memory; LLM generates SQL (insert/select/update/delete) to manipulate stored state. Introduces "chain-of-memory" that decomposes complex problems into sequences of memory operations.
Parameters—
Domainknowledge-graphagent-memoryrag-retrieval
Open SourceYes
PaperView Paper
WebsiteVisit
CodeRepository
sqlsymbolic-memorychain-of-memorystructured

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
72.534p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
72.430p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
72.658p
HotpotQA
77.990p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
73.487p
Sources:ChatDB paper (arXiv:2306.03901); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)ChatDB paper (arXiv:2306.03901); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)ChatDB paper (arXiv:2306.03901); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)ChatDB paper (arXiv:2306.03901); evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)ChatDB paper (arXiv:2306.03901); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)ChatDB paper (arXiv:2306.03901); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)