this post was submitted on 26 Oct 2024
1105 points (97.2% liked)

Science Memes

10923 readers
1901 users here now

Welcome to c/science_memes @ Mander.xyz!

A place for majestic STEMLORD peacocking, as well as memes about the realities of working in a lab.



Rules

  1. Don't throw mud. Behave like an intellectual and remember the human.
  2. Keep it rooted (on topic).
  3. No spam.
  4. Infographics welcome, get schooled.

This is a science community. We use the Dawkins definition of meme.



Research Committee

Other Mander Communities

Science and Research

Biology and Life Sciences

Physical Sciences

Humanities and Social Sciences

Practical and Applied Sciences

Memes

Miscellaneous

founded 2 years ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] uis@lemm.ee 9 points 1 week ago* (last edited 1 week ago) (1 children)

LLMs can't cite. They don't know what a citation is other than a collection of text of a specific style

LLMs can cite. It's called Retrival-Augmented Generation. Basically LLM that can do Information Retrival, which is just academic term for search engines.

You'd be lucky if the number of references equalled the number of referenced items even if you were lucky enough to get real sources out of an LLM

You can just print retrival logs into references. Well, kinda stretching definition of "just".

[–] notthebees@reddthat.com 1 points 1 week ago (2 children)

My question is that the thing they are citing actually exists and if it does exist, contains the information it claims.

[–] uis@lemm.ee 1 points 1 week ago* (last edited 1 week ago)

the thing they are citing actually exists

In case of RAGs it exists in searched dataset.

and if it does exist, contains the information it claims.

Not guaranteed.

[–] FutileRecipe@lemmy.world 1 points 1 week ago

Depends. In my experience, it usually does exist. Now there are hallucinations where GPT makes up stuff or just misinterprets what it read. But it's super easy to read the GPT output, look at the cited work, skim works for relevance, then tweak the wording and citing to match.

If you just copy/paste and take GPT's word for it without the minimal amount of checking, you're digging your own grave.