The Graph Club: HippoRAG, Rethinking Long-Term Memory for LLMs
This Week’s Paper
HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models
Most RAG systems retrieve passages independently. That works well until answering a question requires connecting information scattered across multiple documents.
HippoRAG asks a different question: what if LLM memory worked a little more like human memory?
Inspired by how the hippocampus helps us recall related information, HippoRAG combines LLMs, knowledge graphs, and Personalized PageRank to connect concepts and retrieve relevant information across multiple hops.
The paper shows that this approach can outperform strong RAG baselines on multi-hop question answering while performing the reasoning-heavy retrieval in a single step.
What we’ll discuss:
Why standard RAG struggles when information is spread across documents
How HippoRAG turns text into a graph of connected knowledge
Why Personalized PageRank is useful for retrieval
Where HippoRAG performs well and where it still falls short
What this could mean for long-term memory in AI systems
You do not need a neuroscience background or deep familiarity with the paper. If you work on RAG, agents, knowledge graphs, memory, or LLM infrastructure, there should be plenty to dig into.
The discussion will be led by Aditya Srivastava, AI Researcher & Engineer at HydraDB and an alumnus of Carnegie Mellon University and IIT Bombay.
Who Should Attend?
AI researchers, ML engineers, data and infrastructure engineers, graduate students, and builders working on RAG, knowledge graphs, agent memory, graph machine learning, or LLM reasoning.
About The Graph Club
The Graph Club is a paper-reading community for researchers and engineers who want to stay close to emerging AI research, strengthen their understanding of foundational ideas, and connect academic work to the systems being built today.
About HydraDB
HydraDB is a fast graph database built on object storage, purpose-built for modern AI workloads (think agent memory, company brain, ontologies etc.). We make AI stateful.
Read the paper before the session:
https://arxiv.org/pdf/2405.14831