Back to Arena

LightRAG

by HKUDS (HKU Data Intelligence Lab)

System Card

OrganizationHKUDS (HKU Data Intelligence Lab)
Released2024-10
Architecturegraph-rag / Dual-level entity/relation retrieval
DetailsBuilds a KG from entity-relation extraction, then retrieves at two granularities: low-level (specific entities) and high-level (broader concepts), combined with vector search. Supports Neo4j, PostgreSQL, MongoDB backends.
Parameters—
Domainrag-retrievalknowledge-graph
Open SourceYes
PaperView Paper
CodeRepository
graph-ragdual-levelentity-extractionemnlp-2025academic

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
2/5
RULER
70.746p
NIAH
no data
LooGLE
no data
LongBench
603p
∞Bench
no data
Multi-Turn Recall
0/2
LoCoMo
no data
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
73.837p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
71.949p
HotpotQA
71.953p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
73.282p
Sources:LightRAG paper (arXiv:2410.05779); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)LightRAG paper (arXiv:2410.05779); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)LightRAG paper (arXiv:2410.05779); evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)LightRAG paper (arXiv:2410.05779); evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)LightRAG paper (arXiv:2410.05779); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)LightRAG paper (arXiv:2410.05779); evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)