Visualizers
Learn by moving things
Interactive explainers for the ideas behind retrieval, embeddings, and agents. A formula tells you what is true; dragging a vector shows you why. Each one also appears inside the article it belongs to, and runs entirely in your browser, so drag, poke, and break it.
How much of the stack is on-topic?
From: How to Judge a RAG System →Context precision: of the retrieved chunks, how many are actually relevant, and are the relevant ones ranked first. The same chunks can score 0.81 or 0.19 depending only on their order, because the model reads and weights the top chunks most.
Four chunks come back, ranked. Three are relevant (green), one is noise (grey). Precision asks how much of the stack is on-topic.
Did it leave a needed page on the shelf?
From: How to Judge a RAG System →Context recall: of the facts the correct answer needs, how many were actually retrieved. Even one missing fact means a silently incomplete answer, and precision (which only grades what came back) cannot see it.
The correct answer needs 5 distinct facts. Recall asks how many of them retrieval actually brought back.
Did the answer stick to the book?
From: How to Judge a RAG System →Faithfulness: of the claims in the generated answer, how many are supported by the retrieved context. A claim that is true in the world but not in the context still counts as a hallucination, because the system invented rather than grounded it.
The generated answer makes 5 claims. Faithfulness checks each one against the retrieved context.