Top all-time · #18213
Polymathic
Company blog · polymathic-ai.org/blogaiscience
Polymathic is the company blog at polymathic-ai.org, writing about ai and science. Since 2023, 4 different Hacker News readers have linked it 4 times. No feed was found, so read it on the site itself.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link Polymathic AI's xVal post, which describes a method for handling numbers in language models as continuous values instead of tokens. Commenters point to it as a more interesting alternative to the lab's announcement.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- /blog/xval/2 people · 2 mentions · first linked 2023-10-10
What Hacker News says?
Tbh the announcement is fine, but this paper is probably more interesting: https://polymathic-ai.org/blog/xval/
Comment linking /blog/xval/
on “Polymathic AI”Submitted Polymathic AI as a story · 9 points
HN discussionNeat! > This is probably due to the way larger numbers are tokenised, as big numbers can be split up into arbitrary forms. Take the integer 123456789. A BPE tokenizer (e.g., GPT-style) might split it like: ‘123’ ‘456’ ‘789’ or: ‘12’ ‘345’ ‘67’ ‘89’ One of the craziest LLM hacks that doesn't get love is https://polymathic-ai.org/blog/xval/ xVal basically says "tokenizing numbers is hard: what if instead of outputting…
Comment linking /blog/xval/
on “Apple: Embarrassingly Simple Self-Distillation Improves Code Generation”Submitted Walrus: A Cross-Domain Foundation Model for Continuum Dynamics as a story · 2 points
HN discussion