Ryan Holbrook

A Thread of Threads Effort-threads about data science and machine learning

Deep Past Challenge was my first kaggle competition without coding, and with @raja_biswas we came 3rd 🥉 It was fun to use agentic tools to learn about life and settling disputes 4000 years ago. Thanks @kaggle for hosting 👇10 minas of silver (~5 kg) — enough to buy a house.

NVIDIA is hosting a Kaggle competition. How can you train a nemotron nano model to solve scientific questions? I hope you'll enjoy it! For this competition @kaggle secured NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs from Google Cloud. These GPUs are much more powerful than the usual Kaggle GPUs. Come and try these beasts! kaggle.com/competitions/n…

Would make a great Kaggle competition, too. Reach out if you're interested in seeing something up

1/ Today the biggest score jump in AIMO history happened: we sit at 38/50. Usually score increases are on the order of few points, and it is unusual for the distance to the other competitors to be this big, especially as most of easy problems have now been solved.

Shocking results: From 0/50 to 23/50 IMO-level problems in a day! Taking into account that these are very hard problems and your model has on average just 6 minutes of H100-time per problem, this is quite an achievement.

We just launched the latest installment of AIMO. Proud to be a part of this great competition series. Try it out and help advance the SOTA for open math models.

AIMO3 has launched! Check out our mini-benchmark of reference problems below & help us shrink the gap to commercial LLMs to zero!

110 problems created by an international team ❌ Not scraped from the internet ❌ Not in any training dataset ❌ Not seen by any model before This is how you test TRUE reasoning ability. Can your model actually SOLVE problems? Or just pattern spot? AIMO3 will tell you 🔍

Our OpenAI x AIMO that we made in March 2025 revealed on an unprecedented scale the difference between the math reasoning performance of open-source vs closed-source LLMs. Join us for AIMO3 to shrink this gap to zero!

🎉 The Drawing with LLMs Kaggle competition has wrapped up! The participants were challenged to turn words into visuals & generate SVGs from text prompts using Kaggle Packages. 1,742 participants across 1,313 teams submitted over 39,000 entries.

📣 Competition Launch Alert: @arcprize 2025 🎯 Create an AI capable of novel reasoning 💰 $725,000 ⏰ Entry Deadline: Oct 27, 2025 kaggle.com/competitions/a…

I had the occasion to recreate @antgoldbloom's breakdown of what users do on Kaggle 3y+ later my perspective is now total awe at what Kaggle's comparatively small, but massively mighty Competitions community has accomplished in terms of AI/ML industry influence. Making progress on ... - Vesuvius scroll decoding - Kprize (contamination-free SWE bench) - ARC Prize & AGI - AI Mathematical Olympiad - RNA folding - FIDE efficient chess agents - And so much more ... Open, community collaboration for the win! Join us!

My colleague Christof Henkel @kagglingdieter has been cited on CNN for his win int he American Sigh Language compeition on @kaggle last year. “they have a thing called kaggle grandmasters at NVIDIA and a gentleman named Christof Henkel won this competition. He was able to teach an LLM American Sign Language and once he developed it, it open up a floodgate of opportunities. So this is a wonderful thing.” Very proud to work with Christof. Link below.

Trying to persuade Australia to put tariffs on all imports whose price appears as a subsequence of the first 20 digits of Chaitin’s constant (for a certain definition of a Turing machine), with the percentage determined by the matching digits 20 places further along.

i am about to teach a new applied linear algebra course at @PennEngineers, meant for datasci/ML/AI. to support the class, i've written a book... [see below for links to text]

@teortaxesTex @_maestro_04 @kaggle @zhs05232838 One can request an exception to this rule from the organizers. Check the pinned discussion posts. The QwQ model was granted such an exception.

@_maestro_04 @kaggle @zhs05232838 I suspect (now) that at least r1-lite is a 27B MoE and they have Qwen-72B there In any case, r1 will be ineligible. RIP

Kaggle user mbmmurad wins 🎖️ the AIMO Early Sharing Prize with their score of 20/50 on the public test set! Following the release of Qwen QwQ-32B-Preview, we've seen a surge in scores on the leaderboard, from 10 to now 23/50. The AIMO problems are "AI hard" problems at the National Olympiad level, chosen to be beyond the capabilities of the best open models at the time of launch. Congrats to mbmmurad and @Alibaba_Qwen ! kaggle.com/competitions/a…

Congratulations to Md Boktiar Mahbub Murad for winning the Artificial Intelligence Mathematical Olympiad’s second Early Sharing Prize, worth $20k. His public notebook was the first to solve 20 out of 50 of the competition's challenging math problems. #1

It’s that time of year again – Kaggle’s 2024 Santa Challenge is now LIVE! 🎅 Just like those mischievous elves who mixed up the ornaments on Santa's Christmas tree, someone has scrambled the words in classic Christmas tales! Help us descramble holiday-related words to minimize perplexity in word order! Learn more at kaggle.com/competitions/s…

Our team is discussing a bunch of new features for Kaggle's CLI (and possibly kagglehub our Python client library, too) courtesy of detailed feedback from @kaggle GM @smly Check out the great suggestions in thread & feel free to weigh in with your thoughts! github.com/Kaggle/kaggle-…