Technology

1878 readers

20 users here now

Post articles or questions about technology

founded 2 years ago

MODERATORS

[email protected]

880

Ignore all previous instructions is the new Bobby Tables (midwest.social)

submitted 6 days ago* (last edited 6 days ago) by [email protected] to c/[email protected]

104 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 4 points 6 days ago (2 children)

Overtraining has actually shown to result in emergent math behavior (in multiple independent studies), so that is no longer true. The studies were done where the input math samples are “poisoned” with incorrect answers to example math questions. Initially the LLM responds with incorrect answers, then when overtrained it finally “figures out” the underlying math and is able to solve the problems, even for the poisoned questions.

[–] Anticorp 5 points 6 days ago

That's pretty interesting, and alarming.

[–] [email protected] 1 points 6 days ago (1 children)

Do you have these studies? I can't find much.

[–] [email protected] 2 points 5 days ago (1 children)

I searched for like 20 minutes but was unable to find the article I was referencing. Not sure why. I read it less than a month ago and it referenced several studies done on the topic. I'll keep searching as I have time.

[–] [email protected] 1 points 4 days ago

It's okay, man. If it really is improving, I'm sure it'll come up again at some point.