TechTakes

1787 readers

112 users here now

Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.

This is not debate club. Unless it’s amusing debate.

For actually-good tech, you want our NotAwfulTech community

founded 2 years ago

MODERATORS

[email protected]

LLMs average <5% on 2025 Math Olympiad; award each other 20x points (arxiv.org)

submitted 1 week ago by [email protected] to c/[email protected]

44 comments fedilink hide all child comments

"Notably, O3-MINI, despite being one of the best reasoning models, frequently skipped essential proof steps by labeling them as "trivial", even when their validity was crucial."

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

[–] [email protected] 28 points 1 week ago (32 children)

"thought process" lol.

[+] [email protected] -16 points 1 week ago (30 children)

LLMs are a lot more sophisticated than we initially thought, read the study yourself.

Essentially they do not simply predict the next token, when scientists trace the activated neurons, they find that these models plan ahead throughout inference, and then lie about those plans when asked to say how they came to a conclusion.

[–] [email protected] 26 points 1 week ago* (last edited 1 week ago) (11 children)

You didn't link to the study; you linked to the PR release for the study. This and this are the papers linked in the blog post.

Note that the papers haven't been published anywhere other than on Anthropic's online journal. Also, what the papers are doing is essentially tea leaf reading. They take a look at the swill of tokens, point at some clusters, and say, "there's a dog!" or "that's a bird!" or "bitcoin is going up this year!". It's all rubbish dawg

[–] [email protected] 18 points 1 week ago (1 children)

To be fair, the typesetting of the papers is quite pleasant and the pictures are nice.