Back to Arena

HippoRAG 2

by OSU NLP Group

System Card

OrganizationOSU NLP Group
Released2025-02
Architecturegraph-rag / Extended PPR with deeper passage integration
DetailsBuilds on HippoRAG's Personalized PageRank algorithm with deeper passage integration and more effective online LLM use. Positions RAG as a non-parametric continual learning mechanism outperforming standard RAG on factual, sense-making, and associative memory tasks.
Parameters—
Domainrag-retrievalknowledge-graphlifelong-learning
Open SourceYes
PaperView Paper
CodeRepository
icml-2025continual-learningkgassociative-memory

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
74.846pEstimated
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
8392pEstimated
Multi-Hop QA
3/3
BABILong
79.484pEstimated
MultiHop-RAG
7579pEstimated
HotpotQA
75.574pSelf-Reported
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
76.597pEstimated
Sources:arXiv:2502.14802 Table 2 — F1 with Llama-3.3-70B-Instruct backboneArena estimate — derived from capability profile, not independently verified