Digest | Garden
All Bookmarks 3566

@AmishMagic

Like everyone, I've been a bit distracted exploring @deepseek_ai R1 and experimenting with it...

Jan 31

@Avichal Garg (Electric Capital) ⚡

1/ This screenshot is floating around and I know the back story.

Jan 31

@Om Alve

So I pre-trained an LLM from scratch this week, yes, pretrained not finetuned. This is how it went...

Jan 31

@Wordgrammer

Okay. Thanks for the nerd snipe guys. I spent the day learning exactly how DeepSeek trained at 1/30...

Jan 29

@Noahdgoodman

Congrats to OAI on producing a reasoning model! Their opaque tweets demonstrate that they’ve...

Jan 29

@Abhinav Upadhyay

End-to-end ML School was a wonderful learning resource. Probably one of the earliest tutorials on...

Jan 29

@N8 Programs

reading a deepseek paper and stumbled upon a very beautiful formula where they *unify SFT and MOST...

Jan 29

@Zach Wilson

You only need to read four books to truly get what’s going on in ML and data engineering:

Jan 29

@Edward Beeching

As part of our open reproduction of R1, we have roughly reproduced DeepSeek's MATH-500 eval numbers...

Jan 29

@Jasmine Sun

1st image is my notes for an essay on LLMs and conversational interfaces. notice how rough they are

Jan 28

@Elio

is RL seriously just weighing logprobs of an action by the reward, its that simple? why so much...

Jan 28

@Chuck Smith

@signulll I'm running the 70b version on a 2023 Mac Pro (M2 Ultra, 192 GB unified memory). It runs...

Jan 28

@Ben Smith

Fasting is literally the greatest remedy—the doctor within.

Jan 27

@Taoki

they really brought the whole squad out

Jan 26

@Freeman Jiang

life update: i'm taking a gap term and moving to new york city to pursue independent research at...

Jan 26

@Matheson Mayotte ✝️

Matheson Mayotte ✝️: @CohenSite

Jan 23

@Arnaud Bertrand

@gammalab_tweets Speaks to a different vision of AI as a commodity for everyone rather than a...

Jan 23

@Wh

How to train a State-of-the-art reasoner.

Jan 22

@KK👋🏼

This process has yielded close to 99% accuracy with pdf data extraction for me:

Jan 22

@Aravind Srinivas

You can try DeepSeek-R1 on We'll try to bring it up on core Perplexity in the context of advanced...

Jan 22

@Austen Allred

PS you can invest with me

Jan 19

@Austen Allred

We've changed how our AI for Developers training works!

Jan 19

@Austen Allred

Doing final review of this week's GauntletAI apps, and wow I'm impressed.

Jan 19

@MatthewBerman

1/ SakanaAI just dropped their latest research: Transformer²

Jan 19
auto_stories

Select a bookmark to read

Click any item from the list