Aadiv
@crooningmoon.bsky.social
ML enthusiast with a focus on CV looking to get into computational chemistry. Not a bot, I swear!
Reposted by Aadiv
A short list of tips for keeping a clean, organized ML codebase for new researchers: eugenevinitsky.com/posts/quick-...
Eugene Vinitsky
eugenevinitsky.com
December 18, 2024 at 8:00 PM
A short list of tips for keeping a clean, organized ML codebase for new researchers: eugenevinitsky.com/posts/quick-...
Could you elaborate on the benchmaxed model? Shouldn't the deployed model be tested on the benchmarks, as opposed to fine-tuning it for that?
April 7, 2025 at 2:01 AM
Could you elaborate on the benchmaxed model? Shouldn't the deployed model be tested on the benchmarks, as opposed to fine-tuning it for that?
Hey this looks great! Could you please clarify if it's remote and whether international students can apply? Thank you!
December 6, 2024 at 1:27 AM
Hey this looks great! Could you please clarify if it's remote and whether international students can apply? Thank you!
Ironic considering they mentioned this exact case in their o1-pro launch. They did say they were working on their search to mitigate this, has that been implemented already?
December 6, 2024 at 12:50 AM
Ironic considering they mentioned this exact case in their o1-pro launch. They did say they were working on their search to mitigate this, has that been implemented already?
Oh hey that's great to know! For some reason I can't message you on bsky, do you have a discord or something else?
December 5, 2024 at 5:44 PM
Oh hey that's great to know! For some reason I can't message you on bsky, do you have a discord or something else?
What if you speak a niche language but a lead already exists? Can there be multiple leads for a single language?
December 5, 2024 at 1:18 PM
What if you speak a niche language but a lead already exists? Can there be multiple leads for a single language?
Reposted by Aadiv
Did you know that attention across the whole input span was inspired by the time-negating alien language in Arrival? Crazy anecdote from the latest Hard Fork podcast (by @kevinroose.com and @caseynewton.bsky.social). HT nwbrownboi on Threads for the lead.
December 1, 2024 at 2:50 PM
Did you know that attention across the whole input span was inspired by the time-negating alien language in Arrival? Crazy anecdote from the latest Hard Fork podcast (by @kevinroose.com and @caseynewton.bsky.social). HT nwbrownboi on Threads for the lead.
Seconding this, anything by Rick Riordan really
December 1, 2024 at 12:50 PM
Seconding this, anything by Rick Riordan really
I've noticed this peculiarity with deepseek also, sometimes it'll just switch over to chinese as well!
November 28, 2024 at 4:38 AM
I've noticed this peculiarity with deepseek also, sometimes it'll just switch over to chinese as well!
Might just be me but I just start drooling when I see stuff like "Llama 4 training on 100k H100s" JUST ONE OF THOSE COULD CHANGE MY LAB'S TRAJECTORY
November 25, 2024 at 5:43 PM
Might just be me but I just start drooling when I see stuff like "Llama 4 training on 100k H100s" JUST ONE OF THOSE COULD CHANGE MY LAB'S TRAJECTORY
Could you elaborate on it being a consequence of 10043?
November 21, 2024 at 1:13 PM
Could you elaborate on it being a consequence of 10043?
My initial assessment is it's pretty good at focused logic, but falls apart a bit when it comes to teaching or simplifying code. But hey, for 50 free prompts a day who cares?
November 21, 2024 at 11:20 AM
My initial assessment is it's pretty good at focused logic, but falls apart a bit when it comes to teaching or simplifying code. But hey, for 50 free prompts a day who cares?
If there was ever an artist I'd want working with stuff like this it's Jacob. Brilliant work!
November 21, 2024 at 10:32 AM
If there was ever an artist I'd want working with stuff like this it's Jacob. Brilliant work!
I'm not too sure about this, after ~2 hours of experimenting I find it much harder to prompt it to think in a certain way, or explain its reasoning in a way I understand. While this may be more powerful out of the box, it definitely needs more work done at the interpretability front.
November 21, 2024 at 5:13 AM
I'm not too sure about this, after ~2 hours of experimenting I find it much harder to prompt it to think in a certain way, or explain its reasoning in a way I understand. While this may be more powerful out of the box, it definitely needs more work done at the interpretability front.
An amazing primer on how to save yourself from the heartbreak of that desk reject
For those who missed this post on the-network-that-is-not-to-be-named, I made public my "secrets" for writing a good CVPR paper (or any scientific paper). I've compiled these tips of many years. It's long but hopefully it helps people write better papers. perceiving-systems.blog/en/post/writ...
Writing a good scientific paper
perceiving-systems.blog
November 21, 2024 at 2:04 AM
An amazing primer on how to save yourself from the heartbreak of that desk reject
Reposted by Aadiv
A paper a day keeps the FOMO away (mega-thread).
In the thread below, you'll find my daily ramblings about research articles I particularly enjoyed. Let's see how long I can keep this up!
In the thread below, you'll find my daily ramblings about research articles I particularly enjoyed. Let's see how long I can keep this up!
November 15, 2024 at 10:12 PM
A paper a day keeps the FOMO away (mega-thread).
In the thread below, you'll find my daily ramblings about research articles I particularly enjoyed. Let's see how long I can keep this up!
In the thread below, you'll find my daily ramblings about research articles I particularly enjoyed. Let's see how long I can keep this up!