Back to Arena

HippoRAG

by OSU NLP Group (Ohio State University)

System Card

OrganizationOSU NLP Group (Ohio State University)
Released2024-05
Architecturegraph-rag / Hippocampal-inspired KG with Personalized PageRank
DetailsUses an LLM as neocortex to transform a corpus into a schemaless knowledge graph serving as an artificial hippocampal index. Given a query, HippoRAG identifies key concepts and runs Personalized PageRank on the KG to integrate information across passages, achieving single-step multi-hop retrieval.
Parameters—
Domainrag-retrievalknowledge-graphlifelong-learning
Open SourceYes
PaperView Paper
CodeRepository
hippocampusknowledge-graphpagerankmulti-hopneurips-2024

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Data Transparency:1 self-reported5 estimated
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
74.142pEstimated
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
73.535pEstimated
Multi-Hop QA
3/3
BABILong
70.39pEstimated
MultiHop-RAG
7579pEstimated
HotpotQA
59.29pSelf-Reported
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
70.858pEstimated
Sources:arXiv:2405.14831 Table 4 — F1 with IRCoT+HippoRAG (ColBERTv2); EM 45.7. HippoRAG alone: F1 55.0Arena estimate — derived from capability profile, not independently verified