The leading platform for crowdsourced benchmarks, competitions, and hackathons where the community shapes real-world progress in AI.
San Francisco
Joined October 2009
- Tweets5.7K
- Following298
- Followers322K
- Likes3.3K
Kaggle retweeted
Got a 5th place solo gold in the recent @kaggle @scrollprize competition which was about tracing the layers of 2,000-year-old scrolls buried by Vesuvius so they can be digitally unrolled and read. It’s a bummer I dropped on private LB, but I am really proud of the custom SDF loss! 📜🔥
Kaggle retweeted
Check out this great work on measuring progress towards AGI and the associated global @Kaggle hackathon. I continue to believe that Minimal AGI will be achieved in the coming years: an AI that can do all the cognitive things that people can typically do.
How do we measure progress toward AGI? It takes a village – and a bit of healthy competition. 🛠️
We’re launching a global hackathon with @Kaggle to build new cognitive evaluations for AI. With $200k in prizes up for grabs, help us put our framework to the test.
Join the challenge → goo.gle/4uqOVKK
Kaggle retweeted
.@GoogleDeepMind is introducing a new cognitive framework for measuring progress toward AGI. Build and run your own evaluations against frontier models using @Kaggle’s Community Benchmarks.
Join the “Measuring Progress Towards AGI: Cognitive Abilities” hackathon: goo.gle/40CvS2r
Kaggle retweeted
How do we measure progress toward AGI? It takes a village – and a bit of healthy competition. 🛠️
We’re launching a global hackathon with @Kaggle to build new cognitive evaluations for AI. With $200k in prizes up for grabs, help us put our framework to the test.
Join the challenge → goo.gle/4uqOVKK
Kaggle retweeted
Help us measure the progress towards AGI (specifically cognitive capabilities) by building benchmarks on @kaggle, with $ 200K in prizes available! Details in 🧵
Kaggle retweeted
The State of Machine Learning Competitions 2025 by @ml_contests is out, and as always it gives a great snapshot of how ML competitions are changing and what actually worked in winning solutions last year.
Some of the interesting takeaways are:
• More than 390 machine learning competitions ran in 2025 across 30+ platforms.
• @kaggle remains the largest platform by users and number of competitions, and many winners still trained on free notebook environments like Kaggle and @GoogleColab .
• Almost all winning teams used Python as their main programming language.
• PyTorch dominates deep learning in winning solutions, with TensorFlow now rare.
• Pandas remains the default DataFrame library, with Polars seeing growing use.
• Gradient-boosted trees remain the standard for tabular problems, often combined with neural models in ensembles.
• Transformers overtook CNNs in computer vision competitions for the first time.
• @Alibaba_Qwen models dominated language competitions.
• Despite earlier claims about Kaggle grandmaster-level autonomous agents, there is still no strong evidence of them succeeding in live competitions.
Kaggle retweeted
Huge thanks to @kaggle, @chauhan_nilay16 in particular for helping us bring Tiny Aya to their platform. Making the model easy to explore and experiment with lowers the barrier for researchers and builders everywhere. 🌏
Kaggle retweeted
Large language models are seemingly great at everything EXCEPT this very basic poker skill…
Feat @DougPolkVids
Kaggle retweeted
🥇 1st Place Secrets: XGBoost Tips from 5x Kaggle Grandmaster Chris Deotte
Want to climb the @Kaggle leaderboards? Our own @cdeotte shared his master list of XGBoost Tips and Tricks to help you build faster, more accurate models.
The Winning Formula:
⚡ GPU-Accelerated: Set device="cuda" for massive speedups in training and iteration.
📊 Native Categoricals: Use enable_categorical=True to boost accuracy without the encoding headache.
🛠️ Feature Engineering: Create interaction features that reveal hidden data patterns.
📉 Memory Efficiency: Scale your models by optimizing data types for GPU VRAM.
Whether you're chasing Kaggle Gold or optimizing production ML, these insights from the world’s top-ranked data scientist are essential.
➡️ Read the guide: kaggle.com/writeups/cdeotte/… 🙌
Kaggle retweeted
Incredible gameplay. The gap between these models is closing fast.
If you want to see the full Poker breakdown (plus the new Chess and Werewolf leaderboards), check out the full stats on the Arena.
GG to the devs. 🤝 kaggle.com/game-arena
Kaggle retweeted
It all comes down to this. 🏆
The best chess AIs, the best werewolf players, and best poker models. Who will take the titles?
We’re looking at the best of the best today, LIVE with @NickSchulman. Don’t miss it. youtube.com/watch?v=vzMj2KOy…
Kaggle retweeted
We're advancing AI benchmarking systems by letting them play games 🕹️
Games like Poker & Werewolf in the @Kaggle Game Arena allow us to test AI capabilities and "soft skills" in controlled sandbox environments before they’re deployed 🧵 (1/4) ↓
The @Kaggle Game Arena has added our results for Werewolf, Poker and Chess.♟️✨
These challenges test the skills AI models need for the real world – like contextual communication, building consensus, and navigating ambiguity.
Find out how Gemini did → goo.gle/4qU4qbS
Kaggle retweeted
Day 1 was wild. We saw models playing the odds and others actually trying to push opponents off hands. Can an LLM actually have a "poker face"? 🤔
Day 2 of the Game Arena AI Showdown starts NOW. Nick and I are breaking down the poker semifinals, werewolf highlights, and to see how well AI has improved at chess.
Come hang out! 👇
youtube.com/watch?v=4TJwlPVj… @kaggle @NickSchulman
Kaggle retweeted
So stay tuned to my X this week, and if you want a deeper dive I'll be interviewing one of the @GoogleDeepMind geniuses behind the project over on the Win-Win YT channel. NO DELVING, I PROMISE.
Excited to be covering this alongside @DougPolkVids @NickSchulman @GMHikaru - thanks @kaggle for including us all!
Kaggle retweeted
The AI field is in need of harder benchmarks to test capabilities of the latest AI models. This update to @Kaggle Game Arena with werewolf and poker (heads-up) plus chess, gives us new objective measures of real-world skills like planning and decision making under uncertainty.
Kaggle retweeted
We know AI dominates the chess board ♟️, but can it also handle a bluff?
Today we’re moving from perfect calculation to also look at calculated risk. I’m LIVE with @NickSchulman for the @Kaggle Game Arena AI Showdown. 🃏
Gemini vs. GPT vs. Claude and more. Let's see who has the guts to go all-in.
🔴 Watch here: youtube.com/watch?v=6rb2rMah…
RT @GoogleDeepMind: We’re adding Werewolf, Poker and new Chess results to the @Kaggle Game Arena.♟️✨
These challenges test the skills AI m…
Kaggle retweeted
Here's to an interesting week with a @kaggle event with @NickSchulman and then Speed Chess Championship in London with @chesscom @AlirezaFirouzja @MagnusCarlsen and Denis Lazavik. Mon, Tues and Wed will be on Kick, Twitch and Youtube for some variety!!
Kaggle retweeted
🥇 3rd out of 3,395 teams in @kaggle Santa 2025 ($10k)
Problem: pack N Christmas tree polygons into a square of minimal side length.
Approach: recursive rectangle composition, heavily modified Sparrow, and C++ SA + push compression.
Write-up below 👇
Kaggle retweeted
I want @kaggle to be a place where anyone can come to learn, build skills, solve high impact challenges, and participate in pushing forward the state-of-the-art in AI.
A huge part of making that a reality is supporting organizations that share this vision. That’s why I’m so thrilled to sponsor Women in Data Science (WiDS) returning to Kaggle for their 9th annual Global Datathon. 🏆
Excited to see the impact this year's challenge holds!
Learn more about the competition ✨ kaggle.com/competitions/WiDS…
ALT WiDS Global Datathon 2026 Predicting Time-to-Threat for Evacuation Zones Using Survival Analysis Overview When a wildfire ignites, emergency managers face an impossible question with incomplete information. Which fires will reach populated areas? How quickly? And which communities should prepare first? This year's WiDS Global Datathon challenges participants to build survival models that answer these questions using only the earliest signals available. Your task is to predict the probability that a wildfire will threaten an evacuation zone within 12, 24, 48, and 72 hours, drawing on data from just the first five hours after ignition. The goal is not a single prediction but a calibrated forecast across multiple time horizons. Emergency responders need both urgency rankings (which fires demand immediate attention) and probability estimates they can trust when making high-stakes decisions about evacuations, resource deployment, and public alerts.