Chunking
Split documents along their own structure, such as headings and paragraphs, and keep a little overlap so a sentence is not cut from its context. Store the source and position with each chunk.
Retrieval
Measure retrieval on its own: for a set of questions with known source passages, how often does the right chunk appear in the top results? If it does not, no prompt will fix the answer.
Citations
Ask the model to answer only from the retrieved passages and to cite them. Then check that each cited passage actually supports the claim.
Next
A full build with code and an evaluation set will follow.
