Skip to main content

Massive data: in-depth summary

816
·
230 chars
·
70 tokens
·
For an in-depth summary, it is recommended to ask two rounds of questions. During the second inquiry, have it refer back to the original text to check for errors or omissions. Excerpted from an article by Esor Huang.
Prompt Content
Based on the content from articles '@1' to '@3', please analyze and identify any errors or inconsistencies that need to be corrected or supplemented. Make sure your analysis is consistent with the original content of the articles.
FAQ
Why two rounds of summarization?
The first round may include AI fabrication or omissions. The second round requires "check for errors and omissions against the original," forcing the AI back to the source. Single summaries drift easily; the double-pass mechanism is a practical trick against AI hallucination.
Does the AI actually find errors on second check?
It finds obvious errors (dates, numbers), but deeper drift (tone, implied stance) is still hard to catch. For more rigor, a third pass asking the AI to "list all key points omitted from the original" surfaces more overlooked details than two passes.
How do I use this prompt?
Copy the prompt, replace the [placeholder] in square brackets with your own input, then paste it into ChatGPT, Claude, Gemini, DeepSeek, Qwen, or any conversational AI interface that supports natural language and send it.
Share
Discussion