LLMs Can Teach Themselves to Better Predict the Future. (arxiv.org)
from Cat@ponder.cat to technology@lemmy.world on 12 Feb 2025 19:30
https://ponder.cat/post/1613097

#technology

threaded - newest

A_A@lemmy.world on 12 Feb 2025 22:03 collapse

The basic model of DeepSeek-R1 14B was already groundbreaking since it reached the level of GPT-1o. But this does much better by bring it to the level of GPT-4o

Authors are from :

1 - Lightning Rod Labs (USA)

…
www.lightningrod.ai/about

2 - (UK)

London School of Economics and Political Science


Machine learning is still developing very fast.
“We used 8, H100 GPUs, for training.”
Huge amounts of processing power are not required.