Home_markov_chain_rag_tutorial

---

type: index up: "WIKI-INDEX"

markov_chain_rag_tutorial

Welcome to the project wiki. This is an LLM-maintained knowledge base that grows as the project evolves.

Navigation

• Index — Full catalog of all wiki pages
• Log — Chronological record of wiki updates

This wiki is the concept companion to the Markov Chain RAG Tutorial. Start with Markov Retrieval Tutorial for the whole argument in one page, then follow the concept links below. The runnable code and the primary writeup (docs/tutorial.pdf) live in the project repository.

Categories

Start here

• Markov Retrieval Tutorial — what this repository is and the three claims it makes

Concepts

• Random Walk With Restart — the retrieval algorithm
• Heterogeneous Graph — the chunk / document / concept substrate
• Path-Length Decomposition — reading relevance by hop count
• Restart Probability — the c dial from local re-ranking to global mixing
• Vanilla RAG Failure Mode — why nearest-neighbor retrieval misses tail vocabulary
• Bridge Entity — how the walk crosses vocabulary clusters
• Extraction-Retrieval Co-Design — why extraction and retrieval are one design problem

The demonstration

• Tutorial Corpus — the 12 entities and 24 chunks
• Tutorial Queries — the five queries and the findings
• Tutorial Figures — the three generated figures

See the Index for the full catalog and SCHEMA for wiki conventions.