←back to thread

LLMs can get "brain rot"

(llm-brain-rot.github.io)
466 points tamnd | 1 comments | 21 Oct 25 14:24 UTC | HN request time: 0.263s | source
Show context
AznHisoka ◴[21 Oct 25 14:30 UTC] No.45656299[source]▶
Can someone explain this in laymen terms?
replies(4): >>45656501 #>>45657077 #>>45658026 #>>45666082 #
PaulHoule ◴[21 Oct 25 14:47 UTC] No.45656501[source]▶
They benchmark two different feeds of dangerous tweets:

  (1) a feed of the most popular tweets based on likes, retweets, and such
  (2) an algorithmic feed that looks for clickbait in the text
and blend these in different proportions to a feed of random tweets that are not popular nor clickbait and find that feed (1) has more of damaging effect on the performance of chatbots. That is, they feed that blend of tweets into the model and then they ask the models to do things and get worse outcomes.
replies(1): >>45657029 #
ForHackernews ◴[21 Oct 25 15:29 UTC] No.45657029[source]▶
Blended in how? To the training set?
replies(1): >>45660602 #
1. PaulHoule ◴[21 Oct 25 19:35 UTC] No.45660602[source]▶
Very early training.