Tag: rag
All the articles with the tag "rag".
-
Neo4j Vector Index — SEARCH was a syntax error, but the warning told me to switch to SEARCH (Studying Graph RAG #2)
I created a vector index in Neo4j and attached embedded equipment descriptions as node properties. I tried searching with SEARCH and got a syntax error, so I went back to queryNodes — but running that produced a warning telling me to switch to SEARCH. In the end, I pulled both vector-matched equipment and its alarms in a single query.
-
Neo4j Cypher — I used RETURN with properties and the graph view wouldn't show (Studying Graph RAG #1)
I decided to study graph RAG and started by setting up a free Neo4j Aura instance. I created nodes and relationships with Cypher, but the query results only showed up as a table, not a graph. It turned out to be about what you pass to RETURN.
-
RAG Embedding Comparison — Measuring recall@k on My Blog Data (OpenAI vs bge-m3)
After establishing the ["feeling-based benchmarking → numeric benchmarking" principle](/en/posts/quant-study-00-pandas) in my quant retrospective, I actually quantified an embedding model comparison this time. I indexed 441 chunks from my blog posts with OpenAI text-embedding-3-small and bge-m3 respectively, then measured recall@3 with a test set of 20 question-answer source pairs. Overall: OpenAI 80% vs bge-m3 90%. bge-m3 hit 100% on hard-difficulty questions — the decisive factor was connecting to the source text by meaning even when words didn't overlap. On easy questions, the misses turned out to be caused by typos (cladue, underscores) — a twist showing the grading criteria itself was wrong.
-
LLM Eval Study #2 — The Pitfalls of Similarity-Based Evaluation, 5 Principles of Test Set Design, and a Misunderstanding About the Term 'Regression Test'
My second session studying Eval. I ran similarity-based evaluation (embedding cosine similarity) myself and got an unexpected result — both OpenAI's text-embedding-3-small and bge-m3 gave the highest score to the 'wrong answer.' This is because similarity captures topical/expressive closeness, not content correctness. Just a difference in markdown formatting can swing similarity scores significantly. In practice, combining similarity with LLM-as-Judge is the standard approach. Also covers 5 principles of test set design, plus a terminology correction: it's not a 'regression test,' it's improvement validation (A/B).