A pile of research notes is not knowledge — it's a pile. The useful move is pulling out the entities and the relationships between them as you go: who or what is involved, and how they connect to each other. That structure is what lets you answer a question later without re-reading everything.
This is the same idea behind knowledge-graph research tools: instead of one flat wall of text, you end up with a small map — nodes and connections — that you can query directly. You don't need special software to start; you need the habit of asking "who's connected to what here" every time you read something.
Run the seed and pull the entities and relationships out of the passage yourself before checking the model's version. Notice what you missed.