BestPromptFinder

Find Why a RAG Answer Missed Something Before You Rewrite the Prompt

Quality, usefulness and confidence are AI evaluations out of 100 — not user ratings. Community results come from visitors' "worked / didn't work" votes below. How scoring works →

The prompt

Help me debug one wrong answer from my retrieval-augmented (RAG) system. Don't rewrite the prompt until we know where the evidence was lost. Question asked: [paste] Answer given: [paste] What the right answer should include: [e.g. the exception in clause 4.2] Document and passage that contains it: [name, section, paste the passage] Retrieved chunks for this question (in rank order, with scores if you have them): [paste] The final prompt sent to the model, including the context block: [paste] Work through these checks in order and stop at the first one that fails: 1. Ingestion: is the passage in the index at all, and was it split so the condition and the exception stayed together? 2. Retrieval: is the right chunk in the top results? If not, why might the query and chunk not match (wording, acronyms, language)? 3. Ranking and limits: was it retrieved but cut off by top-k, a reranker or a token limit? 4. Prompt: was it in the context but ignored? Quote the line in the prompt that should have made the model use it. 5. Answer: did the model use it but summarise away the condition? Output: the failing stage, the evidence for that, and one targeted fix with a test question to confirm it. Don't guess about data you haven't been shown; ask for it.

Why this prompt

• Hand-picked, editorially reviewed prompt
• Written for ChatGPT, Claude, Gemini

Source

Created by BestPromptFinder — editorial prompt, not user-generated. See our content & source policy.

Related Coding prompts