LocalLLaMA

3 readers

1 users here now

Community to discuss about Llama, the family of large language models created by Meta AI.

founded 1 year ago

MODERATORS

communick@poweruser.forum

Ilya paper about Q* (alien.top)

submitted 11 months ago by Xnohat@alien.top to c/localllama@poweruser.forum

16 comments fedilink hide all child comments

Ilya from OpenAI have published a paper (2020) about Q* : a GPT-f model have capabilities in understand and resolve Math, Automated Theorem Proving.

https://arxiv.org/abs/2009.03393

When AI model can understand and really doing Math, that a critical jump.

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

[–] mrpogiface@alien.top 1 points 11 months ago (3 children)

I think that's a bit out of date. My guess is its building on this work

https://openai.com/research/improving-mathematical-reasoning-with-process-supervision

[–] nested_dreams@alien.top 1 points 11 months ago (2 children)

This definitely sounds like the paper. 100% worth the read, surprised I hadn't heard much about it until this ordeal

[–] jack-in-the-sack@alien.top 1 points 11 months ago

I'll definitely be asking my GPT to read this paper to me as my bedtime story.

[–] wind_dude@alien.top 1 points 11 months ago (1 children)

PRM8k, made the rounds maybe 6+ months but they never publicly released the model.

[–] dododragon@alien.top 1 points 11 months ago

I've recently just got into LLM's have you tried these math models? They seem to follow math related instructions reasonably well.

wizard-math:13b-q6_KMathLLM-MathCoder-CL-7B.Q8_0.ggufmetamath-mistral-7b.Q5_K_M.gguf

[–] _Lee_B_@alien.top 1 points 11 months ago

Strange, I thought they would naturally be rewarding the process, by rewarding each word that's generated by the sequence to sequence model, rather than the final words, for example. Maybe they over-optimised and skipped training on all output.

[–] wind_dude@alien.top 1 points 11 months ago

yea, that seems to be what a few news articles have referenced.