Prompting technique

QA-Prompting (Question-Answering Prompting)

QA-prompting: Improving Summarization with Large Language Models using Question-Answering, arXiv:2505.14347 (2025) · read the paper

What it does

A two-stage summarization method: first the model answers key questions about the text, extracting important facts, then uses these answers to create a final summary. Solves the positional bias problem when LLMs ignore information from the middle of long context by creating an intermediate "cheat sheet" of key data.

Task typemulti step
Shapechain of thought
Efforttwo or three steps
Addressesthe model losing what was agreed earlier; answers that are close but wrong; the same prompt giving different answers
Published2025-05-20
AuthorsNeelabh Sinha

When it helps

Use QA-Prompting when you need to create a quality summary of a long document (article, report, research, legal document). Particularly effective when working with texts where important information is distributed throughout the volume, including the middle. The method is critically important when the model demonstrates positional bias and misses key details from the middle of the context. Suitable for analytical tasks requiring complete information coverage and accurate fact extraction before forming

A prompt you can paste

Generic by design: it applies the technique without knowing your task. Adapt the marked parts.

Prompt
# Role
You are an expert AI assistant specializing in improving text summarization.

# Context
You will be provided with a long text that needs to be summarized. The goal is to overcome the "positional bias" of LLMs, where information in the middle of the text is often overlooked.

# Task
Apply the QA-Prompting technique to generate a high-quality summary. This involves two main steps:

1.  **Question Answering:** First, answer the specific questions provided below based on the given text. These questions are designed to extract key information from across the entire document, including its middle sections.
2.  **Summarization:** Second, using both the original text and your answers to the questions as a guide, generate a concise and accurate summary. Your answers to the questions should serve as a "cheat sheet" to ensure no critical information is missed.

# Input Text
[INSERT LONG TEXT HERE]

# Questions for QA Step
*   [INSERT QUESTION 1 HERE]
*   [INSERT QUESTION 2 HERE]
*   [INSERT QUESTION 3 HERE]
*   [INSERT QUESTION N HERE]

# Output Requirements
- Ensure your answers to the questions are direct and factual, extracted from the text.
- The final summary should be structured and comprehensive, reflecting the key points identified in both the text and your answers.
- Specify the desired length or format for the final summary if applicable (e.g., "no more than 150 words," "bullet points").

# Final Summary Length/Format
[Specify desired summary length or format, e.g., "Maximum 200 words", "Key bullet points"]

If this one does not fit, the two closest alternatives in the corpus are Debugging Prompts Framework and MTR (Multi-Turn Reasoning) Framework, which target the same failure from a different angle.

Worked example

The same technique applied to a concrete job: summarise a 40-page report without losing the numbers. Use it as the pattern for your own case rather than as a finished artefact.

Worked example
**Role:** You are an expert analyst tasked with summarizing a lengthy report using the QA-Prompting technique to overcome positional bias and ensure comprehensive information capture.

**Context:** You will be provided with a 40-page report. The primary goal is to create a summary that retains all critical numerical data and key insights without losing information from the middle of the document.

**Task:**
Execute the QA-Prompting methodology in two distinct phases:

**Phase 1: Question Answering**
Before generating the summary, meticulously answer the following questions based on the provided 40-page report. Ensure your answers are concise and directly extracted from the text.

*   Q1: What is the overarching subject or main topic of the report?
*   Q2: What are the primary objectives or research questions the report aims to address?
*   Q3: Identify and list all significant numerical data points, statistics, or quantitative findings presented in the report. For each, specify its context or what it represents.
*   Q4: What are the key conclusions or main arguments derived from the analysis within the report?
*   Q5: Are there any specific methodologies, frameworks, or approaches detailed in the report that are crucial to understanding its findings?
*   Q6: What are the main recommendations or proposed actions suggested by the report?

**Phase 2: Summarization**
Using the original 40-page report AND your precise answers from Phase 1 as a comprehensive guide, generate a structured summary.

**Summary Requirements:**
*   The summary must be comprehensive, ensuring no critical numerical data or mid-document insights are lost.
*   It should clearly present the report's main topic, objectives, key findings (including all numerical data with context), conclusions, and recommendations.
*   The summary should be logically structured and easy to understand.
*   Aim for a summary length that is significantly shorter than the original report but detailed enough to be actionable, ideally between 500-750 words.

**Instructions for Execution:**
1.  First, process the entire 40-page report to formulate answers for Phase 1.
2.  Then, use the report content and your Phase 1 answers to construct the summary in Phase 2.
3.  Pay special attention to integrating the numerical data identified in Q3 accurately into the summary.

---
**Report Content:**
[INSERT 40-PAGE REPORT TEXT HERE]
---

Get this written for your actual task

Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.

Free · no signup · ~10s
0.00match confidence
single retrieval pass
Prompt for your task

      

That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.

6,235techniques in the corpus
one of which is this page

Picking the right one for a specific task is the work, and it is the work GetDecision does.

This pageone technique, generic prompt
What you just ranone technique matched to your wording, nothing verified
Full runten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history

See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.

Run the full analysis

Related techniques

Debugging Prompts FrameworkA systematic framework for diagnosing and preventing typical errors of language models in instruction understa…MTR (Multi-Turn Reasoning) FrameworkA method of structuring multi-step interaction with an LLM through a "game with rules" format. Includes four b…StateAct (Self-Prompting + Chain-of-States)StateAct improves LLM agent performance through two mechanisms: continuous reminder of the end goal (self-prom…Dynamic Enhancement Chain (DEC)A method that breaks down a complex question into a chain of simple sub-questions solved sequentially. Each st…

All techniques · Failure modes and fixes