This one is interesting : https://t.co/XrqKnKbrl7 I think this tries to think beyond simple agent memory and resonates with your LLM wiki idea. @karpathy
I strongly suspect that Claude Mythos is a looped language model, as described in the paper "Scaling Latent Reasoning via Looped Language Models" from ByteDance
The authors of that paper called out graph search as one of the areas where looping provides a huge theoretical advantage over standard RLVR. And look at where Mythos blows out its competitors the most
i might have heard the same π -- I guess info like this is passed around but no one wants to say it out loud.
GPT-4: 8 x 220B experts trained with different data/task distributions and 16-iter inference.
Glad that Geohot said it out loud.
Though, at this point, GPT-4 is probably distilled to be more efficient.
@brianschardt I wonder if the prompt made it correctly to ask for βstock price predictionβ instead of βarticle sentiment analysisβ. For example the Lyft layoff news got a low score of 3, but itβs actually a good news for stock price.
Congrats @InSilicoMeds@biogerontology on this encouraging development: the FIRST in history AI-discovered and designed anti-fibrotic product candidate entering Phase 1 human clinical trial!
https://t.co/734NztkoXa