h a l f b a k e r yOK, we're here. Now what?
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
This Large Language Model interface would read LLM responses, identify highly similar parts of responses, and remove them, outputting only the most salient parts of the original answer. So anything starting with "as an artificial..." "I hope this clears up..." most paragraphs starting with "it's important
to note that..." "While it's true that..." "While I cannot..." "you make an interesting point..." "in summary..." and other common paragraphs which add long and unnecessary verbiage to an answer would be eliminated.
Sentences within these paragraphs containing statistically unusual words and phrases would be preserved and condensed and rephrased into a single paragraph of output.
[link]
|
|
Interestingly to me: of the commonly known LLMs only GPT-4 was willing to even address this. It said:
The Large Language Model Condenser would process responses to identify repetitive or generic phrases and remove them, leaving only the most relevant information. By filtering out common introductory or concluding phrases, and retaining unique words and phrases, the output would be more concise and focused.
The other models all answered with a variant of "I apologize, but I do not feel comfortable manipulating or editing my own responses in the way you have described." |
|
|
Interesting experiment [+] |
|
|
Repeated iterations of this algorithm will tend towards the single most concise answer possible to any general question. |
|
|
^ This, [a1], thank you for this. It's a rainy day here at Camp Teacup, and I'm just waxing nostalgic for our pal Douglas, and Sir Pterry, Robin Williams, Dame Edna, oh, and Rosalind Russell, Lucille Ball, Marjorie Main and Percy Kilbride. |
|
| |