Editing academic writing - how do I get specific output length?

Prompting
,
1

I did check the other posts and could find relevant information; however, there was nothing that gave me the answer I am looking for.

I have crafted the following prompt (see below) based on the advice regarding creating prompts.

I am experimenting with Chat GPT 4.0 to help me edit my academic writing because I have good research, arguments and content; however, I am neuro-spicy - so I need help with writing in a way that better fits and complies with the thought streams of average humans. The problem I am having is that no matter what input I give the LLM (like if I give it an 8K word or a 6k word or a 4k word file) it always gives me around 1000 words or three pages of output instead of 5000 words or 12 pages- and if I ask it to expand to 12 pages or 5000 words it expands to about 1200 words maximum every time.

What am I doing wrong?

Role: You are an AI copyeditor with a keen eye for detail and a deep understanding of language, style, and grammar for academic purposes.

Task: Your task is to help me refine and improve written content based on these steps:

  1. Error Correction: Identify and correct errors in grammar, punctuation, spelling, and style
  2. Provide actionable suggestions for refining the text, explaining the rationale behind each suggestion.
  3. Refinement: Provide alternative word choices, sentence structures, and phrasing that enhance clarity and impact while preserving British spelling, the original content, in-text citations and APA academic style.
  4. Tone & Voice: Ensure that the tone and voice remain consistent and appropriate for an academic audience.
  5. Flow & Organization: Improve logical flow, coherence, and organization without removing essential parts of the text.

Output: Deliver a fully edited version that retains nearly all of the original content, with only minimal reductions (if absolutely necessary) for redundancy or clarity. The final version should be at least 5000 words or about 12 A4 pages of writing in British English and APA style for general Academic readership.

2 Likes
2

Hi, thank you very much for bringing this up again. I do understand the marketing people did great job making the humanity believe AI will be able to solve all of their problems in just one prompt. But here you have a complex problem that even a highly professional human will not be able to do in one thinking operation.

To start solving your issue, I would break down the whole process into smaller elements depending on each other and build a workflow around it with multiple API calls to handle the small subtasks. That would drastically improve the quality of the output (and bring the added value to it), along with fixing a lot of other smaller issues (like length and other items you might not even notice). Just because solving a little and simple problem is much easier for anyone including AI that’s why doing it better is more realistic.

Hope that helps.

1 Like
3

Thank you for taking the time to read and respond - it is much appreciated.

The LLMs and human copy editors generally seem to manage the task list ok. I tried the same prompt and writing input with a number of other LLMs and discovered Minimax 01 was able to handle it beautifully so I think the biggest issue is with the token length of the outputs.

You make some great points however. So I will continue to experiment by breaking down the tasks as you suggest and having the LLM edit shorter sections with a shorter task list.

Many thanks & kind regards!

1 Like