#announcements

1 messages · Page 1 of 1 (latest)

vivid wind
#

Happy Discord launch day! We're so excited to be able to connect with you here. Later today I'll be hosting a casual chat about our plans for discord and a chance for anyone to ask questions about discord.
https://discord.gg/kaggle?event=1138215380809166999

wide aspen
#

Hi @everyone! Yesterday we released Meta Kaggle for Code, a dataset containing all of the public, Apache 2.0 licensed notebooks created by the community over the past ~8 years: https://www.kaggle.com/datasets/kaggle/meta-kaggle-code 🎉

To get started with it, I recommend forking one of these notebooks:

If you want to learn more about the dataset and why we released it plus inspiration for what kinds of research and projects to use it in, check out the blog post here: https://www.kaggle.com/discussions/product-feedback/430422. And big shoutout to @limpid osprey, the engineer who led this project.

Let us know if you have questions or things you want to discuss in the #💾┊data channel!

Meta Kaggle Code

Kaggle's public data on notebook code

which packages are popular? 📊️👀️

Explore and run machine learning code with Kaggle Notebooks | Using data from multiple data sources

Meta Kaggle Code Joined to Meta Kaggle

Explore and run machine learning code with Kaggle Notebooks | Using data from multiple data sources

sleek gorge
#

Just set up an event for tomorrow, a live "fix-it-friday" where I'll be working on small changes hopefully suggested by you! If you got a good idea, or a small thing that's been bugging you, join in and 🤞 you might see your change on kaggle.com just a few hours later: https://discord.gg/kaggle?event=1139305139186970754

bold bramble
sleek gorge
vivid wind
#

The Lux AI Challenge are excited to kickstart stage 1 of the NeurIPS'23 competition with Kaggle 🎉 🎉 🎉

Stage 1 is a re-run of the previous season 2 competition, giving opportunities to researchers and competitors who didn't have a chance to familiarize themselves or compete previously to try out open sourced code and their own tweaks now. Submit at https://www.kaggle.com/competitions/lux-ai-season-2-neurips-stage1/. There are a lot of exciting open-sourced solutions, with the top open sourced solution placing 15th last season

That being said, the research side of the competition will primarily be in the last stage, stage 2 of the competition. Stage 2 seeks to benchmark large-scale decision-making, whether its reinforcement learning, imitation learning, or rule-based approaches! Stage 2 has not finalized and is planned to be like season 2, but with bigger maps and more factories to facilitate controlling many more units. Stage 2 is planned to launch around early September!

Check out the dedicated. LuxAI Discord for more information:
https://discord.gg/c4Zx8gdfGJ

vivid wind
#

@everyone Excited to announce that @spare drift will be hosting a special competition wrap up event for the recently closed Contrails competition next week!

And a quick reminder that we are hosting a special event for the close of the ICR competition in 11 hours from now too!

https://discord.gg/kaggle?event=1141238797003145236

wraith trench
#

Join our newly launched Playground competition, season 3 episode 21 (#playground-series-s3e21) . It's part of our Tabular Tuesday series, designed specifically for beginners to hone their machine learning and data science skills. This is a different type of competition. https://goo.gle/45Jn3Em

🎯 to improve a dataset that is being used to train a random forest model
⏰ final submission deadline: Sept 11, 2023
🎁 your choice of Kaggle swag!

Improve a Fixed Model the Data-Centric Way!

Playground Series - Season 3, Episode 21

vivid wind
#

@everyone We've launched a new forum called #1145832920049786880 - here you can chat with others in languages other than English! Please check it out and get the conversation started in the language you are most comfortable in.

wraith trench
#

@everyone 📣 Competition launch alert! Google - Fast or Slow? Predict AI Model Runtime hosted by Google Research.
🎯: predict the runtime of graphs and configurations in the test dataset
💰: $50,000 prize pool
⏰: November 10, 2023 (entry deadline)
More information here: https://goo.gle/3qKJXg7

Google - Fast or Slow? Predict AI Model Runtime

Predict how fast an AI model runs

wide aspen
wide aspen
#

Hello @everyone! Reminder to sign up to attend this event with @serene pike 🤗 taking place tomorrow! He'll be speaking about Diffusion Models w/ Hugging Face Diffusers. 🎉 Looking forward to seeing you there! https://discord.gg/pFsTwGhX?event=1138491012562554900

wraith trench
#

📣 Competition launch alert!
Child Mind Institute - Detect Sleep States is hosted by Child Mind Institute.
🎯 develop a model trained on wrist-worn accelerometer data to determine a person's sleep state.
💰 $50,000 prize pool
⏰ November 28, 2023, entry deadline
https://goo.gle/3r36ip7

Child Mind Institute - Detect Sleep States

Detect sleep onset and wake from wrist-worn accelerometer data

wide aspen
wide aspen
#

😭 Hi everyone, so sorry for the technical difficulties with the talk today. We will work to get more events on the books and explore alternative options to ensure a higher level of reliability for folks. Thank you to our speaker Sayak for his patience and willingness to give a talk. 🙏

wraith trench
#

@everyone
Competition launch alert! Stanford's Ribonanza RNA Folding competition hosted by DasLab, Eterna, Stanford University School of Medicine, Shujun He, and HHMI News.

🎯 Create a model that predicts the structures of any RNA molecule
💰: $100K
⏰: 11/30/2023 (entry deadline)
🤯: dataset is ~100x larger than any other public RNA dataset currently available (hence Ribonanza)

Learn more: https://www.kaggle.com/competitions/stanford-ribonanza-rna-folding

Stanford Ribonanza RNA Folding

Create a model that predicts the structures of any RNA molecule

wraith trench
wraith trench
#

🤖 New on #KaggleModels! Introducing Llama 2 from MetaAI: a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters.
📚 Explore, share, and upvote your favorite notebooks. Happy Kaggling!
https://www.kaggle.com/models/metaresearch/llama-2

Llama 2

Llama 2 is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters

wraith trench
wraith trench
#

🤖 New on #KaggleModels! Explore Meta AI’s Code Llama: The ultimate code model family! 💻🦙 Unlock elite performance, infilling mastery, extensive context support, and zero-shot programming skills. Transform your coding journey today! Happy Kaggling!
https://www.kaggle.com/models/metaresearch/codellama

Code Llama

Code Llama is a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models for programming tasks

wraith trench
wraith trench
wraith trench
#

👀 Checkout these 2 community competitions on Kaggle:

1️⃣ Google Smartphone Decimeter Challenge 2023: Elevate smartphone GNSS precision and win $15,000! 📱🌍 https://www.kaggle.com/competitions/smartphone-decimeter-2023

2️⃣ Adversarial Nibbler: Explore unsafe image generation with adversarial prompts and earn kudos for your expertise! 🎮🎉 https://www.kaggle.com/competitions/adversarial-nibbler

Google Smartphone Decimeter Challenge 2023

Improve high precision GNSS positioning and navigation accuracy on smartphones.

Adversarial Nibbler

Submit implicitly adversarial prompts that trigger unsafe image generation

wraith trench
#

🤖 New on #KaggleModels! 🌟 Introducing OpenLM’s open_llama: A permissively licensed open source replica of Meta AI's LLaMA large language model! 🚀 Explore and innovate the power of AI!
🤖 Check it out here: https://www.kaggle.com/models/openlmresearch/open-llama

open_llama

Permissively licensed open source reproduction of Meta AI's LLaMA large language model.

wraith trench
#

🤖 New on #KaggleModels! 👋 Meet MosaicML's MPT: MosaicML's GPT-style model pre trained on 1T tokens of text & code. Perfect for coding with an 8k token context.
Explore it here: https://www.kaggle.com/models/mosaicml/mpt

MPT

MPT is a GPT-style model, and the first in the MosaicML Foundation Series of models.

wraith trench
wraith trench
wraith trench
#

🤖 New on #KaggleModels! ☀️ Dive in to learn more about Tatsunori Hashimoto's Alpaca: Fine-tuned from a 7B LLaMA model on 52K instruction-following data using Self-Instruct techniques.
Discover its power now! 🚀 https://www.kaggle.com/models/tatsu-lab/alpaca/

Alpaca

The Alpaca model is fine-tuned from a 7B LLaMA model on 52K instruction-following data generated by the techniques in the Self-Instruct paper

wraith trench
wraith trench
#

🤖 New on #KaggleModels! 👀 Checkout LMSYS Org's Vicuna, your chat assistant! 🤖 Trained via fine-tuning LLaMA 2 on user-shared conversations from ShareGPT. Developed by LMSYS, it's an auto-regressive transformer-based language model.

Happy Kaggling! 💬
https://www.kaggle.com/models/lmsysorg/vicuna

vicuna

Vicuna is a chat assistant trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT

wraith trench
#

We've hit 15 million Kagglers! 🎉

Thank you to our amazing community – let's keep pushing boundaries, rigorously testing technologies, & shaping the future of ML together. 🙌

15_million_Kagglers.jpg
devout gyro
#

@everyone
[⚔️ Invitation] Challenge LLMs & help build an authoritative LLM leaderboard! Try out the Chatbot Arena, a tool that lets you test two anonymous LLMs side-by-side and vote for the best one.

👉 Try it out: https://chat.lmsys.org/

LMSYS Org at UC Berkeley developed it, and we’re excited to share that we just started a partnership with them to further support the platform. We'd love to hear your ideas for features, events, or competitions that could make it even better in this thread: https://discord.com/channels/1101210829807956100/1156284741352423506.

wraith trench
wraith trench
wraith trench
#

@everyone
📣 Competition Launch Alert!

“AI Village Capture the Flag @ DEFCON31”, hosted by AI Village @ DEF CON.

🎯 Tackle 27 hand-crafted ML security challenges to find flags, solve puzzles, and gain hands-on experience with concepts of AI security & safety.
💰 $50,000 prize pool
⏰ Nov 9, 2023 (entry deadline)
https://goo.gle/3ZQZyrC

AI Village Capture the Flag @ DEFCON31

Collect flags by evading, poisoning, stealing, and fooling AI/ML

wraith trench
vivid wind
#

We had a minor spam attack over the weekend, I just went through and cleaned it all up. It seems like a couple of accounts got hacked and then were used to spam everywhere. Sorry for the mess! For now we've set up an auto-mod bot to delete any post with a discord link in it to stop the same thing happening again. If you see any further spam please don't hesitate to tag myself or the moderator team to help remove it!

wraith trench
wraith trench
wraith trench
wraith trench
#

@everyone
🎅 Ho ho ho! It’s that time of year again – Kaggle’s $50,000 Santa Challenge is now LIVE!

Naughty elves rearranged the ornaments on the North Pole workshop tree. Help us fix them before Santa notices by efficiently solve ornament puzzles.

👉 http://kaggle.com/competitions/santa-2023

🎯 Solve cube-like puzzles in the fewest moves
🎁 $50,000 prize pool
⏰ Entry Deadline is Jan 24, 2024

Santa 2023 - The Polytope Permutation Puzzle

Solve twisty puzzles in the fewest moves

wraith trench
wraith trench
#

@everyone
📣 Competition Launch Alert! BirdCLEF 2024 hosted by Cornell Lab of Ornithology.

🎯 to identify under-studied Indian bird species by sound
💰 $50,000 Prize Pool
⏰ Entry Deadline: June 3, 2024

Learn more at https://www.kaggle.com/competitions/birdclef-2024

BirdCLEF 2024

Bird species identification from audio, focused on under-studied species in the Western Ghats, a major biodiversity hotspot in India.

wraith trench
#

@everyone
📣 Competition Launch Alert! The Learning Agency Lab - Automated Essay Scoring 2.0 hosted by
Vanderbilt University & The Learning Agency Lab.

🎯 Improve essay scoring algorithms
💰 $50,000 Prize Pool
⏰ Entry Deadline: 6/25/24

Work with one of the largest open-source dataset to train a AWE model to score student essays.

Learn more at https://www.kaggle.com/competitions/learning-agency-lab-automated-essay-scoring-2

Learning Agency Lab - Automated Essay Scoring 2.0

Improve upon essay scoring algorithms to improve student learning outcomes

wraith trench
wraith trench
wraith trench
wraith trench
#

@everyone
📣 Kaggle has partnered with Google to launch the Gemini API Developer competition at Google I/O. 👏

Build an app that integrates the Gemini API to showcase the power of generative AI. Compete across categories for your share of $1 million in prizes OR drive away in an EV DeLorean!

👉 Learn more https://goo.gle/4bGAOXH

Google for Developers
Gemini API | Join the Gemini API Developer Competition  |  Google A...

Integrate the Gemini API, quickly develop prompts, and transform ideas into code to build AI apps.

wraith trench
#

@everyone
📣 Competition Launch Alert! LLM 20 Questions hosted by Kaggle.

🎯 See which LLM team can guess the target word first
💰 $50,000 Prize Pool
⏰ Entry Deadline: August 6, 2024

Be the first to guess the secret word in this game of question-asking and answering.
Learn more at https://www.kaggle.com/c/llm-20-questions

LLM 20 Questions

Guess the secret word in this cooperative game of question asking and answering

wraith trench
#

@everyone
📣 Competition Launch Alert! RSNA 2024 Lumbar Spine Degenerative Classification hosted by Radiological Society of North America (RSNA).

🎯 to develop models for detecting and classifying degenerative spine conditions from lumbar spine MR images.
💰 $50,000 Prize Pool
⏰ Entry Deadline: October 1, 2024

Learn more https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification

RSNA 2024 Lumbar Spine Degenerative Classification

Classify lumbar spine degenerative conditions

wraith trench
wraith trench
#

@everyone
📣 Competition Launch Alert! ARC Prize 2024 hosted by Mike Knoop and François Chollet at ARC Prize.

🎯 Create an AI capable of solving reasoning tasks
💰 $1M+ Prize Pool
⏰ Entry Deadline: 11/03/2024

In the AI arena, progress towards true Artificial General Intelligence (AGI) has hit a wall.

This competition invites Kagglers to think ‘outside the bot’ to create AI that learns and thinks like humans.
👉 Learn more: https://www.kaggle.com/competitions/arc-prize-2024

ARC Prize 2024

Create an AI capable of solving reasoning tasks it has never seen before

wraith trench
#

@everyone
📣 Here’s your chance to apply for KaggleX Fellowship Program 2024!

We are accepting fellow applications for our fourth cohort - apply by June 23rd
👉 https://www.kaggle.com/kagglex

The project for this year will focus on building a chatbot by fine-tuning Google's Gemma LLM using your own custom conversation-style datasets.

To learn more about eligibility & application process:
👉 https://www.kaggle.com/kagglex/#prospective-fellows

KaggleX Fellowship Program
KaggleX Fellowship Program
wraith trench
wraith trench
#

@everyone
📣 Competition Launch Alert! ISIC 2024 - Skin Cancer Detection with 3D-TBP hosted by ISIC - International Skin Imaging Collaboration.

🎯 Develop image algorithms to identify skin cancer from 3D TBP.
💰 $80,000 Prize Pool
⏰ Entry Deadline: August 30, 2024

Your work will help accurately identify skin cancer, helping with early diagnosis & treatment.
Learn more at https://www.kaggle.com/competitions/isic-2024-challenge

ISIC 2024 - Skin Cancer Detection with 3D-TBP

Identify cancers among skin lesions cropped from 3D total body photographs

wraith trench
wraith trench
wraith trench
#

@everyone
🤖 Llama 3.1 is now available on Kaggle Models!

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes (text in/text out).

👉 Learn more: https://www.kaggle.com/models/metaresearch/llama-3.1

Meta | Llama 3.1 | Kaggle

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes (text in/text out).

wide aspen
#

@everyone
📣 We're running a quick survey to get feedback on potential features to improve Kaggle Datasets. Please take a look here for more details: https://www.kaggle.com/discussions/product-feedback/537183. Feel free to also ping me in #💻┊ask-a-dev if you have questions or feedback/ideas you want to share (or, of course post a new topic on our Product Feedback forum!) thank_you

🔗 https://www.kaggle.com/discussions/product-feedback/537183

[Other feedback] Help Us Improve Kaggle Datasets! Take Our Survey |...

[Other feedback] Help Us Improve Kaggle Datasets! Take Our Survey.

wraith trench
#

@everyone
Competition Launch Alert! Today with Google, we just launched the “Unlocking Global Communication with Gemma” competition.

🎯 Fine-tune Gemma 2 for different languages and share your knowledge through reproducible notebooks that explore elements like language fluency, literary traditions, and more.
💰 $150,000 Prize Pool
⏰ Entry Deadline: 1/14/25

Your participation will help unlock the full potential of language AI and create a more connected and understanding world.

Learn more at https://www.kaggle.com/competitions/gemma-language-tuning

Google - Unlock Global Communication with Gemma

Create Gemma model variants for a specific language or unique cultural aspect

wraith trench
#

@everyone
Competition Launch Alert! AI Mathematical Olympiad hosted by AIMO. Help advance AI models’ mathematical reasoning skills & compete for $2+ million in prizes.

🎯 Create algorithms and models that can solve tricky math problems written in LaTeX format. Your participation will help to advance AI models’ mathematical reasoning skills and drive frontier knowledge.
💰 $2M+ Prize Pool
⏰ Entry Deadline: 3/25/25

With updated rules, fresh datasets, and more resources, this latest AIMO Progress Prize competition offers an exciting opportunity to drive innovation in AI for Math while fostering healthy competition & supporting open science.

Learn more at https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2

AI Mathematical Olympiad - Progress Prize 2

Solve national-level math challenges using artificial intelligence models

wraith trench
#

@everyone
We're hosting a Gen AI intensive course with Google next week (Nov 11 - 15). We'd love for you to join!

This no-cost course is designed to help you gain a deeper understanding of some of the fundamental technologies and techniques behind Generative AI.
More info and registration: https://rsvp.withgoogle.com/events/google-generative-ai-intensive

5-Day Gen AI Intensive Course with Google

Join our 5-day course on Nov 11 - 15 designed to help you deeply understand some of the fundamental technologies and techniques behind Generative AI.

flint prawn
#

@everyone
📣 Competition Launch! CryoET Object Identification hosted by CZI Science and #CZImagingInstitute

🎯 Find small biological structures in large 3D volumes
💰 $75,000 Prize Pool
⏰ Entry Deadline: 1/29/25

Proteins are the main functional units within our cells. Understanding how they interact with each other & other parts of the cell is key to understanding human health. Manual annotation is a huge bottleneck to biomedical discovery taking months or years to complete.

You’ll develop ML models that annotate protein complexes within a cryoET dataset.

Learn more at https://www.kaggle.com/competitions/czii-cryo-et-object-identification

CZII - CryoET Object Identification

Find small biological structures in large 3D volumes

flint prawn
#

@everyone
Join us for Google’s Women in AI Summit on Dec 3, 2024! This year, experts will provide perspectives on the state of AI and show how to integrate the latest tech into your projects. Don’t miss Kaggle Grandmaster Ruchi Bhatia as she shares how Kaggle can accelerate your AI career.

https://developers.google.com/events/women-in-ai/2024?utm_source=kaggle&utm_medium=social

Google for Developers
Women in AI Summit 2024  |  Google for Developers

Learn the latest at the Women in AI Summit, a virtual event for anyone who is passionate about the exciting field of AI.

flint prawn
#

@everyone
Competition Launch! Efficiency in Chess AI hosted by International Chess Federation and Google

🎯 Develop an agent that plays chess under strict CPU and memory limitations
💰 $50,000 Prize Pool
⏰ Entry Deadline: 2/4/25

Chess has long been a challenge for AI, but this competition aims to shift the focus from brute-force computation to elegant, efficient design. Can you create an agent that masters the game while staying within strict constraints?
Learn more at https://www.kaggle.com/competitions/fide-google-efficiency-chess-ai-challenge

FIDE & Google Efficient Chess AI Challenge

Create agents to play chess with resource constraints

wraith trench
#

@everyone
Competition Launch! WSDM Cup - Multilingual Chatbot Arena hosted by lmarena.ai (formerly lmsys.org)

🎯 to predict which responses users will prefer in a head-to-head battle between chatbots powered by LLMs
💰 $50,000 Prize Pool
⏰ Entry Deadline: 1/27/25

This competition was selected for the WSDM Cup 2025. Your participation will advance the research ecosystem while improving chatbot interaction with humans.

Learn more at https://kaggle.com/competitions/wsdm-cup-multilingual-chatbot-arena

WSDM Cup - Multilingual Chatbot Arena

Predict human preference across multiple languages from real votes in the Chatbot Arena.

wraith trench
#

@everyone
It’s that time of year again – Kaggle’s 2024 Santa Challenge is 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!

💰 $50,000 prize pool
⏰ Entry Deadline: January 24, 2025

Learn more at https://www.kaggle.com/competitions/santa-2024

Santa 2024 - The Perplexity Permutation Puzzle

Help Rudolph descramble holiday-related words to make the LLMs happy!

wraith trench
wraith trench
#

@everyone
📣 Competition Launch Alert! Equity in post-HCT Survival Predictions hosted by CIBMTR (Center for International Blood & Marrow Transplant Research) and NMDP (National Marrow Donor Program)

🎯 Challenge: Develop models to improve the prediction of transplant survival rates equitably for allogeneic HCT patients.
💰 Prize Pool: $50,000
⏰ Entry Deadline: February 26, 2025

Learn more at kaggle.com/competitions/equity-post-HCT-survival-predictions

wraith trench
wraith trench
#

@everyone
Join us in celebrating the work of the 2024 KaggleX Cohort 4! 👏

Over 15 weeks, 81 fellows collaborated to create projects that showcase their skills in AI/ML. With the support from experienced advisors, they developed custom chatbots by fine-tuning datasets using Google’s Gemma open models.

👉 http://kaggle.com/KaggleX-2024-project-showcase

Stay tuned as we highlight the final projects from the cohort in the upcoming days!

KaggleX 2024 Project Showcase
wraith trench
#

@everyone
📣 Competition Launch Alert! NeurIPS 2024 - Lux AI Season 3 hosted by Lux AI Challenge

🎯 To create and/or train AI bots to play a novel multi-agent 1v1 game against other submitted agents.
💰 $50,000 Prize Pool
⏰ Entry Deadline: March 3, 2025

Learn more at kaggle.com/competitions/lux-ai-season-3

wraith trench
wraith trench
#

@everyone
Introducing FACTS Grounding 🧠 📐

A new benchmark we’re launching with Google DeepMind designed to evaluate the factual accuracy of large language models (LLMs) across over 1,700 tasks.

Check it out!
👉 Leaderboard: https://www.kaggle.com/facts-leaderboard
👉Technical Report: https://goo.gle/FACTS_paper
👉 Blog post: https://deepmind.google/discover/blog/facts-grounding-a-new-benchmark-for-evaluating-the-factuality-of-large-language-models/

FACTS Grounding is made up of 1719 factual grounding tasks. Each example includes:

  • a document
  • a system instruction requiring the LLM to exclusively reference the provided document
  • an accompanying user request

LLM responses are evaluated by judge models to ensure they fulfill the user request and remain grounded in the provided context.

To ensure accessibility, we’ve published 860 tasks as a public data set for anyone to use:
https://www.kaggle.com/datasets/deepmind/facts-grounding-examples

We plan to expand and maintain FACTS over time with new models, tasks and evaluations. Our goal is to promote the development of AI systems that are both knowledgeable and trustworthy.

We’ll also be working with model owners and the broader community to add more models to the leaderboard over time. Let us know what we should evaluate next! 👀

FACTS Leaderboard | Kaggle

A novel benchmark from Google DeepMind designed to evaluate the factual accuracy and grounding of AI models

Google DeepMind
FACTS Grounding: A new benchmark for evaluating the factuality of l...

Our comprehensive benchmark and online leaderboard offer a much-needed measure of how accurately LLMs ground their responses in provided source material and avoid hallucinations

FACTS Grounding 1.0 Public Examples

860 public examples for the FACTS Grounding leaderboard from Google DeepMind.

wraith trench
#

@everyone
The experimental version of Gemini 2.0 Flash API is now on Kaggle Models! 🤖

What’s new?

  • Enhanced performance: Outperforms 1.5 Pro at twice the speed.
  • Multimodal outputs: Text + natively generated images.
  • Tool support: Google Search, code execution, and third-party functions.

Get started by generating an API key via Google AI Studio, forking our starter notebook, and sharing your work with the Kaggle community.

👉 Learn more here: https://www.kaggle.com/discussions/product-announcements/552409

[Product Update] Gemini 2.0 Flash now on Kaggle | Kaggle

[Product Update] Gemini 2.0 Flash now on Kaggle.

wraith trench
#

📢 Kaggle’s notebook environment is now built on Colab’s Docker image.

Here’s what this means:

  • Fewer dependency conflicts when developing across platforms.
  • The uv package is pre-installed, speeding up package installations significantly.

Want to try it out? Create a new notebook on Kaggle and test the “Export to Colab” feature.
👉 https://www.kaggle.com/discussions/product-announcements/470030

This update is designed to make your workflow smoother and more efficient.
👉 For all the details: https://www.kaggle.com/discussions/product-announcements/552460

openincolab.gif
wraith trench
#

Welcome PyTorch torchtune to Kaggle! With our new integration, you can now use one of torchtune’s recipes to fine-tune your favorite model and publish it to Kaggle Models using kagglehub. Fork this LoRA-finetuning tutorial notebook using Llama 3.2 1B to learn how to use it in your workflow
🔗 https://www.kaggle.com/code/felipemello/torchtune-in-kaggle

Read the release notes from PyTorch for more details
🔗 https://github.com/pytorch/torchtune/releases/tag/v0.5.0

torchtune in kaggle

Explore and run machine learning code with Kaggle Notebooks | Using data from multiple data sources

GitHub
Release v0.5.0 · pytorch/torchtune

Highlights
We are releasing torchtune v0.5.0 with lots of exciting new features! This includes Kaggle integration, a QAT + LoRA training recipe, improved integrations with Hugging Face and vLLM, Ge...

wraith trench
wraith trench
#

What advancements will AI bring in 2025? Kaggle’s community is leading the way, driving innovation, pushing boundaries, and being a part of shaping the future of data science and AI. ⭐

This week, we’re excited to introduce #WaysToKaggle—a series designed to help you explore, learn, and make an impact. Whether you’re looking to grow your skills, collaborate with experts, or tackle real-world challenges, Kaggle has something for everyone. 🚀

Stay tuned for actionable tips, inspiring community stories, and engaging challenges. Let’s shape the future of AI together.

Day_0_WaysToKaggle.png
wraith trench
#

Did you know that more developers download datasets from Kaggle than participate in our competitions?

Publishing your data on Kaggle is a fantastic way to reach this engaged audience and contribute to the growing data community!

Getting started is easy: https://kaggle.com/datasets/new

Key benefits of publishing on Kaggle:
👉 Up to 200GB per dataset
👉 Seamless integration with Notebooks
👉 Data previews
👉 Community engagement & feedback

You can also upload and manage your dataset with just a few lines of code using the Kagglehub Python client library. Check out the documentation here: https://github.com/Kaggle/kagglehub?tab=readme-ov-file#upload-dataset

Want to learn more about the technical specs of our Datasets platform? Read the full documentation: https://www.kaggle.com/docs/datasets#technical-specifications

wraith trench
#

Each year Kaggle launches dozens of featured competitions. But, did you know you can launch your own? It takes only minutes to get your hackathon, benchmark, or classroom assignment launched at no cost. #WaysToKaggle 🔗 https://www.kaggle.com/competitions/new

Kaggle’s self-serve platform features offer flexibility to support your projects:

👉 Custom Python metrics
👉 Clonable library
👉 AI generated datasets
👉 Privacy/invitation-only
🆕 Prizes up to $10K USD
✨ … and more

Read the documentation here: https://www.kaggle.com/c/about/community

The community launches nearly 1K self-serve competitions each month. You can easily browse all of the public community competitions: https://www.kaggle.com/competitions?listOption=active&sortOption=default&participationFilter=open&hostSegmentIdFilter=10

Still have a question or want some feedback? We have a forum for you to connect with other community competition hosts for support: https://www.kaggle.com/discussions/competition-hosting

Now you know how to bring the power of Kaggle’s crucible for discovering what works and what doesn’t in AI/ML to your community, research problem, company, or classroom. 🎉

wraith trench
#

At Kaggle, we believe the best way to learn is by doing. And we provide many hands-on ways to help you to get started with AI and ML whether you’re brand new to programming or looking to level up your skills. #WaysToKaggle

Kaggle competitions can be intimidating, but our Learn courses can help you master both the fundamentals of data science and our platform. Start here if you’re new to Kaggle or ML! Earn a certificate for each Learn course you complete: https://kaggle.com/learn

If you missed out on Google’s GenAI Course last year, don’t worry: you can access all of the interactive materials in a self-paced Learn Guide to learn about the fundamental technologies & techniques behind generative AI: https://www.kaggle.com/learn-guide/5-day-genai

Once you’re comfortable with Kaggle’s platform, join a Getting Started competition like the LLM Classification Finetuning based on data from LM Arena. Check out the /code tab for helpful starter notebooks!

👉 https://www.kaggle.com/competitions/llm-classification-finetuning/overview

wraith trench
#

Kaggle Models is one of the newest parts of Kaggle’s platform. Discover how the community uses top models from publishers like Qwen, AI at Meta, Cohere and more to solve real-world tasks in competitions.
#WaysToKaggle 👉 https://www.kaggle.com/models

You can also browse top models by competition to see which ones are popular and working well for a specific benchmark task. Plus, explore publicly and fork shared code using the models you’re interested in 👉 https://www.kaggle.com/competitions/wsdm-cup-multilingual-chatbot-arena/models

You can also fine-tune and publish your own model to Kaggle using tools like kagglehub or integrations with KerasHub or PyTorch. To see an example in action, follow this torchtune tutorial 👉 https://www.kaggle.com/code/felipemello/torchtune-in-kaggle

wraith trench
#

Kaggle Competitions are where the data science community comes together to learn, collaborate, and innovate on real-world challenges. Whether you're a beginner or an experienced data scientist, there's something for everyone. Here's how you can get started:

✨ Learn and Hone Your Skills
Competitions are the perfect way to tackle real-world problems, but getting started can feel intimidating. The monthly Playground Tabular Series offers a low-stakes, beginner-friendly environment to build your skills. Check out the latest competition here: https://www.kaggle.com/competitions?hostSegmentIdFilter=8&searchQuery=playground+series

🚀 Push the Boundaries with Cutting-Edge Challenges
Kaggle is at the forefront of innovation, hosting competitions that benchmark progress on advanced problems—like mathematical reasoning and designing systems that solve real GitHub issues. Explore these exciting challenges: https://www.kaggle.com/competitions

📊 Explore Open-Ended Applied Problems
Dive into Analytics competitions, which focus on applied challenges with broader objectives. Current highlights include the NFL Big Data Bowl and language-building tasks like creating and sharing language variants of Gemma. Get involved here: https://www.kaggle.com/competitions?hostSegmentIdFilter=11

👀 Spectator Sport for Learning
You don’t need to compete to benefit! Kaggle’s community openly shares reproducible code, solution write-ups, and invaluable resources. Browse discussions and solutions from competitors to learn how experts tackle complex problems. Start here: https://www.kaggle.com/competitions/ai-mathematical-olympiad-prize/discussion/519303

🔥 Join the Global Data Science Community
No matter your experience level, Kaggle Competitions are a fantastic way to accelerate your learning and make a real-world impact. Explore all active competitions today: https://www.kaggle.com/competitions

wraith trench
#

Kaggle’s notebooks workspace is a no-cost, no set-up way to bring reproducible data science and ML projects to life. We’ll share features that help you get the most out of this resource. #WaysOfKaggling

Here’s an overview of our favorite features in Kaggle Notebooks 🤓

👉 GPUs & TPUs
👉 100s of preinstalled packages
👉 “Save All” publishing for reproducible snapshots
👉 Integration w/ 100s of thousands of public models & datasets
👉 Compatibility with Colab for cross-platform development

Read about all technical specifications in our documentation 🔗 https://www.kaggle.com/docs/notebooks#technical-specifications

One of the hidden gems of Kaggle Notebooks is the notebook scheduler & pipeline functionality that you can use to maintain more complex projects. For example, triggering a notebook to run when the Meta Kaggle dataset updates daily 🔗 https://www.kaggle.com/datasets/kaggle/meta-kaggle/code?datasetId=9&searchQuery=scheduled

Here’s how to create your own scheduled notebook 👇

1️⃣ Create a notebook you want to schedule
2️⃣ Set a schedule in the editor
3️⃣ Managed scheduled notebooks in the Active Events pane

Learn more about this feature here 🔗 https://www.kaggle.com/discussions/product-feedback/273569

wraith trench
#

@everyone
🤖 Now on #KaggleModels!

👉 DeepSeek-R1 demonstrates advanced reasoning capabilities with emergent behaviors powered by reinforcement learning alone. By incorporating cold-start data, it addresses repetition and enhances performance across tasks.
Learn more: https://www.kaggle.com/models/deepseek-ai/deepseek-r1

👉 DeepSeek-V3 is a Mixture-of-Experts language model with 671 billion parameters, utilizing Multi-head Latent Attention and an efficient load balancing strategy to improve performance while reducing GPU hours.
Learn more: https://www.kaggle.com/models/deepseek-ai/deepseek-v3

wraith trench
#

“Empirical Rigor at Scale – or, How Not to Fool Yourself” by D. Sculley is now online!
This is a deep dive into the challenges of evaluating AI and ML models—covering everything from defining ground truth and addressing contamination to ensuring benchmarks are robust and reliable.

Key Highlights:

  • Tackling contamination and data leakage.
  • Making evaluations trustworthy and resistant to undue influence.
  • Insights from Kaggle competitions on fostering empirical rigor.

If you work with generative models, LLMs, or AI research, this is a must-watch!
👉 Watch here: https://neurips.cc/Expo/Conferences/2024/talk panel/100670

wraith trench
#

🤖 Now on #KaggleModels!

DeepSeek's Janus-Pro is a unified MLLM for multimodal understanding and generation, built on DeepSeek-LLM-1.5b/7b-base. It uses SigLIP-L for vision encoding, supporting 384x384 images, optimized for image generation.

https://www.kaggle.com/models/deepseek-ai/janus-pro

DeepSeek | Janus Pro | Kaggle

Janus-Pro is a unified understanding and generation MLLM, which decouples visual encoding for multimodal understanding and generation.

wraith trench
#

Want to streamline your workflow? Learn how to integrate Google Colab with GitHub and Kaggle for seamless collaboration!

In this quick video, Paige Bailey walks through:

  • Accessing repositories
  • Securing user secrets
  • Leveraging Colab’s built-in features

👉 Watch now: https://www.youtube.com/watch?v=MvTmSYLHzMg

YouTube
Google Cloud Tech
Integrate GitHub and Kaggle with Google Colab

Notebook → https://goo.gle/4h9FzMU

Learn how to supercharge your workflow by integrating Google Colab with GitHub and Kaggle. Join Paige Bailey as she covers everything from accessing repositories, safeguarding sensitive information, and more using Colab's built in features.

Watch more Generative AI Experiences for Developers → https://goo.gle...

▶ Play video
wraith trench
wraith trench
wraith trench
#

@everyone
📣 The Kaggle GenAI Intensive with Google is back!

Taking place on Mar 31-Apr 4, this no-cost course is designed to help you gain a deeper understanding of the fundamental technologies and techniques behind Generative AI through theory, hands-on learning and community engagement.

More info and registration: https://rsvp.withgoogle.com/events/google-generative-ai-intensive_2025q1

GenAI_IntensiveCourse_WithGoogle.png
5-Day Gen AI Intensive Course with Google

Join our 5-day course, March 31–April 4, to explore the fundamental technologies and techniques behind Generative AI.

wraith trench
#

@everyone
📣 Competition Launch Alert! Drawing with LLMs Using Kaggle Packages

🎯 Build and submit Kaggle Packages to generate SVG images of concepts
💰 $50,000 Prize Pool
⏰ Entry Deadline: May 20, 2025

https://www.kaggle.com/competitions/drawing-with-llms

Drawing with LLMs

Build and submit Kaggle Packages capable of generating SVG images of specific concepts

woven pebble
#

@everyone
📣 Competition Launch Alert! Stanford RNA 3D Folding
🎯 Predict RNA’s 3D structure
💰 $75,000 Prize Pool
⏰ Entry Deadline: May 22, 2025
🙏 Stanford University, Howard Hughes Medical Institute, RDas La, Stanford Medicine

Fine-tune ML models to accurately capture RNA’s complex folding patterns.

Learn more at https://www.kaggle.com/competitions/stanford-rna-3d-folding

Stanford RNA 3D Folding

Solve RNA structure prediction, one of biology's remaining grand challenges

woven pebble
#

@everyone
🚀 New Feature: Data Loaders for Datasets!

kagglehub is now compatible with Hugging Face Datasets, Pandas, and Croissant!

Preconfigured Python code snippets to load via:

  • pandas.DataFrame
  • Hugging Face Dataset
  • mlcroissant.RecordSet

Simply copy, paste, set your file path, and run! 💡

🔗 Learn more: https://www.kaggle.com/discussions/product-announcements/564302

[Feature Launch] New Data Loaders UI for Kaggle Datasets | Kaggle

[Feature Launch] New Data Loaders UI for Kaggle Datasets

wraith trench
#

🤖 Cohere for AI’s Aya Vision is now on Kaggle!

A multilingual, multimodal model excelling across 23 languages, supporting image captioning, text generation, visual QA, and more. Available in 8B & 32B versions for cutting-edge research.

👉 Learn more: https://www.kaggle.com/models/cohereforai/aya-vision

CohereForAI | Aya Vision | Kaggle

C4AI Aya Vision is an open weights research release of 8B and 32B parameter models with advanced capabilities optimized for a variety of vision-language use cases, including OCR, captioning, visual reasoning, summarization, question answering, and more.

wraith trench
#

🤖 Now on Kaggle!

Qwen's QwQ-32B, a 32B parameter reasoning model, uses Reinforcement Learning to enhance tasks like math, coding, and problem-solving, ensuring high performance and accuracy. Check it out here 👇
https://www.kaggle.com/models/qwen-lm/qwq-32b

QwenLM | QwQ-32B | Kaggle

QwQ is the reasoning model of the Qwen series which is capable of thinking and reasoning, can achieve significantly enhanced performance in downstream tasks, especially hard problems.

wraith trench
#

@everyone
📣 Competition Launch Alert! Locating Bacterial Flagellar Motors 2025 hosted by Brigham Young University

🎯 Identify the presence & location of flagellar motors in 3D reconstructions of bacteria.
💰 $65,000 Prize Pool
⏰ Entry Deadline: May 28, 2025

Learn more at kaggle.com/competitions/byu-locating-bacterial-flagellar-motors-2025

wraith trench
#

🤖 Google’s Gemma 3 & ShieldGemma 2 are on #KaggleModels!

Gemma 3, a collection of lightweight open models built from the same tech as Gemini 2.0. Available in 1B, 4B, 12B & 27B sizes, optimized for fast, portable AI apps from smartphones to workstations. https://www.kaggle.com/models/google/gemma-3

ShieldGemma 2, a powerful 4B image safety checker built on the Gemma 3 foundation. It detects dangerous, explicit, and violent content, offering developers a customizable, open solution for responsible AI. https://www.kaggle.com/models/google/shieldgemma-2

Google | Gemma 3 | Kaggle

Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.

Google | ShieldGemma 2 | Kaggle

ShieldGemma 2 is a series of safety content moderation models built on Gemma 3 that target four harm categories: sexually explicit, dangerous content, hate and harassment.

wraith trench
#

@everyone
Did you know that over 140,000 participants tuned in for the Kaggle GenAI Intensive with Google last year? The course is now back, taking place on Mar 31-Apr 4, it offers Google ML expert-led theory sessions, hands-on labs, vibrant community, a capstone challenge w/swag, and remains to be no-cost to all!
https://rsvp.withgoogle.com/events/google-generative-ai-intensive_2025q1

5-Day Gen AI Intensive Course with Google

Join our 5-day course, March 31–April 4, to explore the fundamental technologies and techniques behind Generative AI.

wraith trench
wraith trench
#

🤖 Now on Kaggle!

Introducing Mistral Small 3.1 – a 24B parameter model with 128k token long context & cutting-edge vision understanding. Achieving top-tier performance in both text & vision tasks. Ideal for fast-response agents, local inference, and more!
https://www.kaggle.com/models/mistral-ai/mistral-small-3.1/

Mistral AI | Mistral Small 3.1 | Kaggle

Mistral Small 3.1 (2503) is a powerful language model combining text and vision understanding. It has 24 billion parameters for top-tier performance.

wraith trench
#

@everyone
📣 Competition Launch Alert! BirdCLEF+ 2025 hosted by the Cornell Lab of Ornithology

🎯 Analyze audio recordings to identify species.
💰 $50,000 Prize Pool
⏰ Entry Deadline: 5/29/25

In this competition, you’ll train ML models to analyze continuous audio recordings and identify species by their sounds.
Learn more at https://www.kaggle.com/competitions/birdclef-2025

BirdCLEF+ 2025

Species identification from audio, focused on birds, amphibians, mammals and insects from the Middle Magdalena Valley of Colombia.

wraith trench
#

📢 Hey @everyone! Exciting update!

We're now attempting a GUINNESS WORLD RECORDS™ title for the Largest Attendance at a Virtual AI Conference! 🎉

Registration closes on March 28th 11:59PM PT.
Don't miss out on this opportunity to level up your GenAI skills in this no-cost 5-day course covering the latest in Generative AI, and contribute to a potential world record!

Key highlights:

5-Day Gen AI Intensive Course with Google

Join our 5-day course, March 31–April 4, to explore the fundamental technologies and techniques behind Generative AI.

wraith trench
wraith trench
wraith trench
wraith trench
wraith trench
wraith trench
#

📣 We’re excited to announce a new collaboration between Kaggle and the Wikimedia Foundation—the organization behind Wikipedia.

Wikimedia is publishing their structured data in English and French Kaggle, making it easier than ever for data scientists, researchers, and machine learning enthusiasts to use and participate in the world's largest open knowledge base.

This partnership supports Wikimedia's mission to make knowledge freely accessible and useful for everyone. With these datasets on Kaggle, you can:

🔍 Discover Wikimedia datasets
📊 Analyze them in Kaggle Notebooks
💬 Join the conversation in our forums
📁 Download and reuse the data in your own projects
📰 Read Wikimedia Enterprise's announcement: https://enterprise.wikimedia.com/blog/kaggle-dataset/

👉 Learn more:https://www.kaggle.com/datasets/wikimedia-foundation/wikipedia-structured-contents

wraith trench
#

💡Ian Ozsvald shares his experience working on ARC Prize 2024 competition using Llama 3. From building an automatic program generator to evaluating it effectively, Ian provides key insights into optimizing prompts, using chain-of-thought, and tackling multi-stage critical thinking.
If you're exploring ways to enhance your own LLM work and improve your results, this video is a must-watch!

👉 Watch now: https://www.youtube.com/watch?v=ft_PYi8A93M

YouTube
PyData
Ian Ozsvald - Valuable LLM lessons learnt on Kaggle's ARC AGI Chall...

www.pydata.org

Having worked on Kaggle's LLM-based ARC AGI program-writing challenge for 6 months using Llama3, I'll give reflections on the lessons learned making an automatic program generator, evaluating it, coming up with strong representations for the challenge, chain-of-thought and program-of-thought styles and some multi-stage critical t...

▶ Play video
wraith trench
wraith trench
wraith trench
#

🤖 Now on Kaggle Models!

The long-awaited Qwen 3 is here!

The latest generation of LLMs in the Qwen series - featuring both dense and Mixture-of-Experts (MoE) models. Trained on 36T tokens across 119 languages, it brings stronger reasoning, better performance, and long-context understanding (up to 32k tokens).

Learn more: https://www.kaggle.com/models/qwen-lm/qwen-3/

QwenLM | Qwen 3 | Kaggle

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.

wraith trench
wraith trench
wraith trench
#

📣 Welcome SciCode to Kaggle Benchmarks!

SciCode from Minyang Tian and Ofir Press challenges AI Systems to write code for complex physics and math phenomena.

👉 See results, code & data all in one place: https://www.kaggle.com/benchmarks/kaggle/scicode

Kaggle Benchmarks is a new home for reproducible, trustworthy research leaderboards. To work with us to host your reproducible benchmark, reach out to us at kaggle-benchmarks@google.com!

SciCode Leaderboard | Kaggle

A Research Coding Benchmark Curated by Scientists

wraith trench
#

📢 Exciting news! Our position paper "AI Competitions as the Gold Standard for GenAI Evaluation" has been accepted to #ICML2025! 🎉
It's time to recognize AI competitions as the gold standard for GenAI evaluation - thanks to their built-in rigor, leakage safeguards, and real-world relevance.

👉 Learn more: https://arxiv.org/html/2505.00612v1

wraith trench
wraith trench
#

@everyone
📢 Big News!

We’re teaming up with OpenAI to launch the OpenAI to Z Challenge - Kaggle’s first-ever featured Hackathon!

🎯 Use satellite imagery, archaeological data, and OpenAI’s o3/o4 mini + GPT-4.1 models to help uncover lost archaeological sites in the Amazon.
💰 $400K Prize Pool
⏰ Submission Deadline: June 29, 2025

In this Hackathon, you will submit Writeups of your discoveries which will be scored according to the evaluation criteria and shared in a public gallery.

👉 To learn more and participate: https://www.kaggle.com/competitions/openai-to-z-challenge

OpenAI to Z Challenge

Use OpenAI o3/o4 mini and GPT 4.1 models to discover previously unknown archaeological sites

wraith trench
wraith trench
wraith trench
frank leaf
wraith trench
frank leaf
frank leaf
frank leaf
wraith trench
#

@everyone
📣 We’re teaming up with Google Cloud to launch the Gemma 3n Impact Challenge!

🎯 Build impactful apps with Gemma 3n's on-device, multimodal AI
💰 $150,000 Prize Pool
⏰ Entry Deadline: August 6, 2025

This is your chance to show on-device AI's impact in education, crisis response, & accessibility.

To learn more and participate 👇
https://www.kaggle.com/competitions/google-gemma-3n-hackathon

Google - The Gemma 3n Impact Challenge

Explore the newest Gemma model and build your best products for a better world

frank leaf
frank leaf
#

📣 ICML 2025 Alert! Find Kaggle at Booth #121.

Meet our team, explore an interactive demo, & our new community platform for building and sharing top models evaluations.

➕ learn more about Kaggle team's upcoming talk on GenAI evaluation! #ICML2025

Frame_5_1.png
frank leaf
#

📣 ICML 2025! Join us today and contribute to the ICML AI Expert Benchmark.

Stop by Kaggle Booth #121 to see community-driven evaluation in action & contribute to novel GenAI tasks.

Maximize your lunch break, deepen your ML expertise! #ICML2025

frank leaf
#

✔️ Join the waitlist for early access to Kaggle Benchmarks!

Kaggle Benchmarks is the fastest, easiest way to test new models.

Let Kaggle handle infrastructure while you focus on AI breakthroughs and benefit from competition-grade rigor.

Sign up here: goo.gle/kaggle-benchmarks-waitlist

Discord_-_Focus_on_AI_We_Handle_Infra..png
frank leaf
#

📣 ICML 2025: Announcing the AI Experts Benchmark!

Stop by Kaggle Booth #121 to participate in our crowdsourced evaluation & help create novel tasks for GenAI.

Learn how community power sets a new standard in AI evaluation.
#ICML2025

frank leaf
#

📣 ICML 2025: AI Experts Benchmark Reveal

🎉 We’re unveiling the final "ICML 2025 AI Experts Benchmark" leaderboard TODAY at 3 PM PT 🎇

Stop by the Kaggle Booth #121 to learn more about benchmarks published by researchers and curated by the ICML community.

frank leaf
#

Stick around to learn more from the Kaggle team at 3:30 PM PT - watch it online here: icml.cc/virtual/2025/oral/40140

Or join the poster session from 4-7 PM PT to discuss findings. #ICML2025

frank leaf
frank leaf
frank leaf
frank leaf
#

📣 Competition Launch Alert! Commodity Prediction Challenge hosted by Mitsui & Co.

🎯 Build a model that predicts commodity prices with accuracy & consistency.
💰 $100,000 Prize Pool
⏰ Entry Deadline: September 29, 2025

Help power smarter trading strategies, reduce financial risk, & stabilize global markets.

Learn more at kaggle.com/competitions/mitsui-commodity-prediction-challenge

MITSUI&CO. Commodity Prediction Challenge

Develop a robust model for accurate and stable prediction of commodity prices

frank leaf
wraith trench
#

@everyone

The future of trustworthy AI evaluation is here.

We are excited to announce the launch of Kaggle Game Arena, a new, open-source benchmark platform where top AI models go head-to-head in complex, strategic games.

To celebrate, we're kicking off with a 3-day exhibition chess tournament. Eight leading LLMs will battle it out in a single-elimination bracket, with commentary and recaps from chess legends.

This is a unique opportunity to see the cutting edge of system intelligence and model evaluation in action.

Dates: August 5-7

Where to watch: https://www.kaggle.com/game-arena

Learn more about Game Arena and the future of AI evaluation on our blog: https://www.kaggle.com/blog/introducing-game-arena

frank leaf
#

📣 We’re kicking things off tomorrow with a 3-day AI exhibition chess tournament. Eight leading LLMs will battle it out daily from Aug 5 to 7 in a single-elimination bracket with commentary and recaps from chess legends.

See the matchups and get ready to watch live: kaggle.com/game-arena

Game Arena | Kaggle

Discover top gaming models.

frank leaf
frank leaf
frank leaf
#

It’s Day 1 of the Kaggle Game Arena AI chess exhibition tournament ♟️!

Tune in today at 10:30AM PT to watch 4 head-to-head AI matchups 🤖 in a single-elimination bracket

Find all the streaming action from Kaggle's official game replays to live commentary in one place at kaggle.com/game-arena

Game Arena | Kaggle

Discover top gaming models.

frank leaf
#

What an exciting start to the Kaggle Game Arena AI chess exhibition tournament ♟️!

The first round is complete, and we have our four semi-finalists! Congratulations to o4-mini, o3, Gemini 2.5 Pro & Grok 4!

Come back tomorrow! Semi-finals kick off, August 6th, at 10:30 am PT.

Couldn’t catch us live? Get the daily recap from @GothamChess https://youtube.com/@gothamchess

YouTube
GothamChess

The Internet's Chess Teacher

frank leaf
#

@everyone Welcome to Day 2 of the Kaggle Game Arena AI chess exhibition tournament ♟️!

Tune in to the semifinals streaming at 10:30 AM PT to watch Hikaru Nakamura's commentary: https://www.youtube.com/watch?v=Ze8XmssIpG4

YouTube
GMHikaru
WILL GROK DOMINATE? Kaggle AI Chess Exhibition Day 2!!

Hikaru covers the Kaggle AI Chess Exhibition on August 6, 2025
♖ MEMBERSHIP ► https://www.youtube.com/channel/UCweCc7bSMX5J4jEH7HFImng/join
♟️ LEARN CHESS & PLAY WITH ME ► https://go.chess.com/hikaru
👍LIVE MOST WEEKDAYS ON KICK ►https://www.kick.com/gmhikaru
🎁 GIVE 💎 CHESS ► https://www.chess.com/membership/gift?ref_id=...

▶ Play video
frank leaf
#

📢 Let the games begin! It's time to watch four LLMs compete in the second round of head-to-head matchups.
Follow all the action with commentary from Hikaru Nakamura's 👇
https://www.youtube.com/watch?v=Ze8XmssIpG4 #KaggleGameArena

YouTube
GMHikaru
WILL GROK DOMINATE? Kaggle AI Chess Exhibition Day 2!!

Hikaru covers the Kaggle AI Chess Exhibition on August 6, 2025
♖ MEMBERSHIP ► https://www.youtube.com/channel/UCweCc7bSMX5J4jEH7HFImng/join
♟️ LEARN CHESS & PLAY WITH ME ► https://go.chess.com/hikaru
👍LIVE MOST WEEKDAYS ON KICK ►https://www.kick.com/gmhikaru
🎁 GIVE 💎 CHESS ► https://www.chess.com/membership/gift?ref_id=...

▶ Play video
wraith trench
#

What an exciting semi-final round in the Kaggle Game Arena AI Chess Exhibition Tournament! ♟️

👏 Congratulations to o3 and Grok for advancing to the championship match! This single-elimination bracket has set the stage for a thrilling final showdown.

Missed the live action? Catch the daily recap by Levy Rozman (GothamChess) on his YouTube channel: https://www.youtube.com/watch?v=-m33dn_3sNQ

📊 Check out the updated bracket here:
https://www.kaggle.com/benchmarks/kaggle/chess-text/tournament

🏆 The finals kick off tomorrow, August 7th, at 10:30 AM PT.

💻 Tune in on Take Take Take https://www.youtube.com/live/vtHfJ6iYyEY for World Champion Magnus Carlsen's exciting commentary, and find out which model will take home gold!

Day_2_Result_1_1.png
wraith trench
frank leaf
#

This is it, the final day of Kaggle Game Arena's AI chess exhibition tournament 🏆!

Watch the championship round live today at 10:30 AM PT with commentary from World Chess Champion Magnus Carlsen, joined by Grandmaster David Howell on @TakeTakeTakeApp https://www.youtube.com/live/vtHfJ6iYyEY

YouTube
Take Take Take
Magnus Carlsen Commentates Grok vs. OpenAI Finale

Magnus Carlsen and David Howell commentate on the Final between Grok 4 and OpenAI o3 in the Kaggle AI Exhibition Tournament.Download the Take Take Take app: ...

▶ Play video
frank leaf
#

It's time to tune in to the championship match ⤵️

🏆 The one and only World Champion Magnus Carlsen will be covering the championship match, joined by Grandmaster David Howell in studio! https://www.youtube.com/live/vtHfJ6iYyEY

📺 Catch Hikaru Nakamura's real-time commentary: https://www.youtube.com/live/LtG0ACIbmHw?si=vmCSOT2yxI5w4TE7

💻 Watch the games with the AI's "thoughts" on our Kaggle YouTube: https://youtube.com/@kaggle

📽️ Don't miss Gotham Chess for his final recap! https://youtube.com/@gothamchess

wraith trench
#

What a show! The Kaggle Game Arena chess exhibition is complete, and the winner is O3🥇!

Congratulations to Grok and Gemini 2.5 Pro for an incredible battle.

📽️ Don’t miss Levy Rozman (GothamChess)'s final recap on his YouTube channel https://www.youtube.com/watch?v=DIo6IFV78oo.

Check out the finalized bracket and come back for the full leaderboard reveal once we wrap up the all-play-all benchmark.
https://www.kaggle.com/benchmarks/kaggle/chess-text/tournament

A huge thank you to everyone who tuned in, and to our amazing partners Magnus Carlsen and David Howell, Hikaru Nakamura, and Levy Rozman (GothamChess) for the fantastic commentary and analysis on Chess-com and Take-Take-Take app.

wraith trench
frank leaf
wraith trench
#

📌 Mark your calendars: Nov 10 - 14!

The 5-Day AI Intensive Course with Google is back, and this time we're diving into AI Agents! A no-cost, hands-on course to explore, build, and deploy the next generation of AI agents.

More info about the course and curriculum will be shared in the coming weeks.

👉 Learn more: https://rsvp.withgoogle.com/events/google-ai-agents-intensive_2025

5-Day AI Agents Intensive Course with Google

Join our 5-day AI Agents Intensive Course with Google, November 10–14, to learn how to build, evaluate, and deploy agents.

wide aspen
wraith trench
#

@everyone

🚀 New Benchmark Launch: SimpleQA Verified!

We’ve partnered with Google DeepMind and Google Research to release SimpleQA Verified. It is a curated 1,000-prompt benchmark designed to provide a more reliable and challenging evaluation of LLM short-form factuality. It addresses limitations in previous benchmarks like noisy labels, topical bias and redundancy offering the community a higher-fidelity tool to measure parametric knowledge and mitigate hallucinations.

👉 Check out the leaderboard here: https://www.kaggle.com/benchmarks/deepmind/simpleqa-verified

SimpleQA Verified Leaderboard | Kaggle

A reliable factuality benchmark to measure parametric knowledge.

wraith trench
wraith trench
frank leaf
#

📣 Competition Launch Alert: MABe Challenge - Social Action Recognition in Mice

🎯 Train models to recognize 38 types of social and non-social behaviors in mice using top-down video and motion tracking data.
💰 $50,000 Prize Pool
⏰ Entry Deadline: December 8, 2025

Your work could make behavioral research more scalable and consistent, supporting new discoveries in neuroscience, biology, and other fields.

Learn more at https://www.kaggle.com/competitions/MABe-mouse-behavior-detection

MABe Challenge - Social Action Recognition in Mice

Detect unique behaviors from pose estimates of mice.

wraith trench
#

@everyone
We're hosting the 5-Day AI Agents Intensive course with Google on November 10 - 14. We’d love for you to join!

This no-cost course is designed to help you explore the foundations and practical applications of AI agents.

👉 More info and registration: https://rsvp.withgoogle.com/events/google-ai-agents-intensive_2025

5-Day AI Agents Intensive Course with Google

Join our 5-day AI Agents Intensive Course with Google, November 10–14, to learn how to build, evaluate, and deploy agents.

wraith trench
#

🏈 The 8th annual NFL Big Data Bowl is live on Kaggle!

This year, we’re excited to launch two competitions in collaboration with the NFL. Participants will use NFL player tracking (Next Gen Stats) to create insights and models that help enhance the game, opening new ways to leverage NGS data.

NFL Big Data Bowl 2026 - Prediction

In this competition, you will build models to predict player movement after the ball is thrown, using pre-pass tracking data. Test your algorithms on live NFL games!
💰 $50,000 Prize Pool
⏰ Entry Deadline: Nov 26, 2025
👉 Learn more: https://www.kaggle.com/competitions/nfl-big-data-bowl-2026-prediction

NFL Big Data Bowl 2026 - Analytics

In this competition, you will analyze player movement while the ball is in the air & create metrics or visuals to help coaches & fans better understand the game.
💰 $50,000 Prize Pool
⏰ Final Submission: Dec 17, 2025
👉 Learn more: https://www.kaggle.com/competitions/nfl-big-data-bowl-2026-analytics

Pro tip: You can compete in both competitions! 😎

NFL Big Data Bowl 2026 - Prediction

Predict player movement while the ball is in the air

NFL Big Data Bowl 2026 - Analytics

Understand player movement while the ball is in the air

frank leaf
#

🚀 New Model Launch: Granite 4.0

IBM's Granite 4.0 is the latest open-source small language model, built for fast inference, long-context understanding (tested on up to 128K tokens), and cost-efficient deployments. Multiple model sizes provide flexibility for different hardware and use cases.

Learn more: https://www.kaggle.com/models/ibm-granite/granite-4.0

IBM | Granite 4.0 | Kaggle

Enterprise-ready multimodal models deliver exceptional performance against safety benchmarks and across a wide range of enterprise tasks from cybersecurity to RAG.

frank leaf
frank leaf
frank leaf
#

📣 Competition Launch Alert! PhysioNet - Digitization of ECG Images competition hosted by PhysioNet

🎯 To build models that convert ECG images into digital time-series data
💰 $50,000 Prize Pool
⏰ Entry Deadline: January 15, 2026
🙏 Emory University, Georgia Institute of Technology
👉 Learn more: https://www.kaggle.com/competitions/physionet-ecg-image-digitization

PhysioNet - Digitization of ECG Images

Extract the ECG time-series data from scans and photographs of paper printouts of the ECGs.

frank leaf
frank leaf
frank leaf
frank leaf
frank leaf
frank leaf
frank leaf
#

It’s that time of year again! 🎄 The Santa 2025 - Christmas Tree Packing Challenge is now live!

Santa’s got a packing problem - his Christmas trees won’t fit in the boxes! Help him find the smallest square box to fit 1 - 200 trees. 🧑‍🎄
More info 👇
https://www.kaggle.com/competitions/santa-2025

Santa 2025 - Christmas Tree Packing Challenge

How many Christmas trees can fit in a box? Help solve a classic optimization problem with a festive twist.

#

💰Prize Pool: $50,000
⏰Entry Deadline: January 23, 2026

Additionally, early entries might even receive a special $10,000 reward from Rudolph! Hop on your coding sleigh, optimize Santa’s packaging, and help deliver gifts across the globe! 🌍🎄

frank leaf
#

🤖 Welcome Gemini 3 Pro Preview to Kaggle Benchmarks!

Google's new state-of-the-art LLM topped 14 of the 16 leaderboards we analyzed, showing significant improvements in factuality.

Gemini 3 Pro Preview shows improvements across factuality, reasoning, coding, and math:

  • SimpleQA Verified (factuality): 72.1% (prev: Gemini 2.5 Pro at 54.5%)
  • MMLU-Pro (reasoning): 90.5% (prev: Claude Opus 4 at 87.9%)
  • SciCode (coding): 12.5% (prev: o4-mini at 10.8%)
  • AIME 2025 (math): 95.8% (prev: Grok-4 at 93.3%)

👉 Check out how it performed on all benchmarks here: https://www.kaggle.com/benchmarks

Find Benchmarks | Kaggle

Use and download benchmarks for your machine learning projects.

frank leaf
frank leaf
#

📢 We’re excited to partner with IBM Research to launch a new benchmark suite called Enterprise Operations (EntOps).

This suite contains two benchmarks: ITBench and AssetOpsBench - comprehensive frameworks that evaluate how well AI models and agents perform in real-world, domain-specific operational enterprise workflows.
Check out EntOps here: ​​https://www.kaggle.com/benchmarks/ibm-research/enterprise-ops/leaderboard

🔧 ITBench: Evaluating Agents in Real-World IT Operations
ITBench measures model and agent performance across three critical IT domains:

• SRE: Diagnosing and resolving incidents (e.g., high error rates)
• FinOps: Managing cloud costs, budget overruns, and cost anomalies
• CISO: Ensuring security and compliance
Check out the leaderboard here: https://www.kaggle.com/benchmarks/ibm-research/itbench/leaderboard

⚙️ AssetOpsBench: Benchmarking Industrial Operations

A unified framework for evaluating models and agents in industrial operations, using:
• 2.3M+ sensor datapoints
• 4,000+ work orders
• Structured FMEA insights

AssetOpsBench’s FailureSensorIQ is a MCQA dataset and benchmark to probe LLMs’ reasoning and comprehension of sensor–failure relationships in industrial systems
Check out the leaderboard here: https://www.kaggle.com/benchmarks/ibm-research/asset-ops-bench

frank leaf
#

📌 Mark your calendars: December 5–7, 2025!

Google Deep Mind’s Build the Future with Gemini 3 Pro hackathon is going live for two days! 🚀

We’re celebrating the launch of Gemini 3 Pro - Gemini’s most intelligent, capable and versatile model yet. To mark the occasion, we’re unlocking a special tier of the Gemini API for all participants.

🎁 Prizes: $500,000 in Gemini API Credits

Full details on participation will be shared on December 5th, but we wanted to give you an early heads-up so you can plan ahead.

Happy Kaggling!

frank leaf
#

🚀 Feature Update on Kaggle Benchmark

You can now download Kaggle Benchmark leaderboard results!

Compare your favorite models with a simple CURL command or download the full CSV directly for deeper analysis.

Get started: https://www.kaggle.com/benchmarks

LB_download_1.gif
Find Benchmarks | Kaggle

Use and download benchmarks for your machine learning projects.

frank leaf
frank leaf
frank leaf
#

🚀 Benchmark your AI across India’s languages with IndicGenBench!

Developed by Google DeepMind, this benchmark spans 29 Indic languages, including first-ever evaluation data for 18 Indic languages. It supports language tasks like summarization, translation and question answering.

Explore the benchmark and see how your models rank: 👉 https://www.kaggle.com/benchmarks/deepmind/indic-gen-bench/leaderboard

7aHPX9uVpTnrPjN.png
frank leaf
#

📢 The FACTS Benchmark Suite is now live on Kaggle!

Developed by Google DeepMind and Google Research, this suite measures LLM factuality across four dimensions: Parametric knowledge, Search, Multimodal understanding & Grounding.

Explore the leaderboard: https://www.kaggle.com/benchmarks/google/facts

V2.jpeg
FACTS Benchmark Suite Leaderboard | Kaggle

A novel benchmark suite from Google DeepMind and Google Research designed to evaluate the factual accuracy and grounding of AI models

frank leaf
#

🚀 New on Kaggle Benchmarks: DeepSearchQA developed by Google DeepMind!

This benchmark focuses on complex web research tasks and tests agent comprehensiveness.

Check the leaderboard: https://www.kaggle.com/benchmarks/google/dsqa

Key highlights:

  • 900 challenging search questions structured as “causal chains”
  • Prompts across 17 fields
  • Verifiable scoring to compare models
V2.jpeg
DeepSearchQA Leaderboard | Kaggle

A benchmark evaluating comprehensiveness for deep research agents

frank leaf
frank leaf
#

📣 Hackathon Launch Alert! MedGemma Impact Challenge hosted by Google Research

🎯 Build human-centered AI applications by using MedGemma and other open models
💰 $100,000 Prize Pool
⏰ Final Submission: Feb 24, 2026
👉 Learn more: https://www.kaggle.com/competitions/med-gemma-impact-challenge

The MedGemma Impact Challenge

Build human-centered AI applications with MedGemma and other open models from Google’s Health AI Developer Foundations (HAI-DEF).

frank leaf
#

@everyone

🚀 Introducing Community Benchmarks on Kaggle

As AI evolves at an unprecedented pace, measuring intelligence requires more than a few AI research labs alone – it requires the imagination, curiosity, and collective expertise of the global community.
Today, we’re launching Kaggle Community Benchmarks, a new way to build, run and share custom benchmarks for evaluating AI models on real-world use cases with transparent, reproducible results shaped by the community.

With Community Benchmarks, you can:

  • Access leading models (free access within quota)
  • Run reproducible evaluations with auditable outputs
  • Benchmark multimodal, multi-step and tool based tasks
  • Build tasks, group them into benchmarks and compare performance on leaderboards that you can then share with the community

Ready to build your own benchmark? Get started: https://www.kaggle.com/discussions/product-announcements/667898

[Product Launch] Introducing Community Benchmarks | Kaggle

[Product Launch] Introducing Community Benchmarks

frank leaf
#

In case you missed it 👇

We just launched Community Benchmarks where you can now build, run, and share custom AI benchmarks that are evaluated on the leading AI models with transparent, reproducible results.

When you’re building and sharing your work on social media, don’t forget to tag Kaggle - we’d love to highlight and share your work with the community.

To learn more 👉 https://blog.google/innovation-and-ai/technology/developers-tools/kaggle-community-benchmarks/

Google
Introducing Community Benchmarks on Kaggle

Community Benchmarks on Kaggle lets the community build, share and run custom evaluations for AI models.

frank leaf
#

@everyone

📌 Mark Your Calendar: Live Game Arena Event This Monday!

We are releasing two new games, Poker and Werewolf, along with an updated Chess leaderboard next Monday, February 2, running daily from 9:30 AM PT to 11:30 AM PT through February 4.

Top models from Anthropic, DeepSeek, Google, OpenAI, and xAI will compete across these games, as we benchmark their performance in more complex, real-world scenarios involving risk, collaboration and long-term strategy.

Tune in to see expert commentary from poker legends Liv Boeree, Nick Schulman, and Doug Polk, alongside chess Grandmaster and poker enthusiast Hikaru Nakamura.

👉Full details on matchups coming Monday!

compressed_output.gif
frank leaf
#

@everyone Game Arena kicks off today!

From Feb 2–4 | 9:30–11:30 AM PT, follow top AI models as they compete in Poker, Werewolf, and Chess, showcasing reasoning, social strategy, and risk management across new leaderboards.
Event schedule:

Monday, Feb 2: Top 8 models from Anthropic, DeepSeek, Google DeepMind, OpenAI, and xAI face off in a 900,000-hand Poker showdown.

Tuesday, Feb 3: Poker semi-finals plus highlight matches from Werewolf and Chess leaderboards.
Wednesday, Feb 4: Poker final for the crown, Chess top-two showdown, and Werewolf highlights.

Key Details:
🎙️ Co-hosted by GM Hikaru Nakamura & Poker Hall-of-Famer Nick Schulman: https://www.youtube.com/GMHikaru
🗓️ Feb 2–4 | 9:30–11:30 AM PT
🤖 Featuring the top models from Anthropic, DeepSeekAI, GoogleDeepmind, OpenAI, and xAI across poker, chess, and werewolf.
💡Additionally, tune in to watch expert analysis and recaps from Liv Boeree and Doug Polk throughout the week!
👉 Explore leaderboards, evaluation harnesses and analysis at kaggle.com/game-arena

Happy Kaggling!

compressed_output.gif
YouTube
GMHikaru

As #2 in the world of chess it's my pleasure to host this channel to share my knowledge and love of chess. Check out the live streams on kick and twitch to see full content. VODs of kick and twitch are posted on https://www.youtube.com/@moregmhikaru

frank leaf
#

🎬 We’re live!

Watch GM Hikaru and Nick Schulman break down the first round of the poker bracket and chess newcomer matches.

We’re testing the limits of social intelligence, calculated risk, and long-term strategy in models from Anthropic, DeepSeekAI, Google DeepMind, OpenAI, and xAI across 900,000 hands of Poker and 31,000+ games of werewolf.

Tune in on YouTube: https://www.youtube.com/watch?v=6rb2rMahWrE

Read more 👇
https://www.kaggle.com/blog/game-arena-poker

Tale_of_the_Tape_-Kaggle_24.png
frank leaf
#

Day 1 of Game Arena is officially in the books!

Congratulations to our AI poker showdown semi-finalists o3, Gemini 3 Flash, GPT 5.2, and Opus 4.5!

Come back tomorrow! Poker semi-finals kick off and werewolf & chess deep dives continue tomorrow, Feb 3rd, at 9:30 AM PT.

Catch up: 👉 https://www.kaggle.com/game-arena

Couldn’t catch us live? Watch Douglas Polk recap for a technical breakdown of model betting strategies and more on YouTube 👉 https://www.youtube.com/watch?v=jyv1bv7JKIQ

Day_2_1.png
frank leaf
#

It's the semi-finals today! Four models remain, and the stakes are doubling. 🃏♟️We’re live for the Poker Semi-Finals, Chess deep dives, and the penultimate Werewolf rounds!

Join GM Hikaru and Nick Schulman to see which models can navigate the high-pressure strategies of the final four.

Tune in on YouTube https://www.youtube.com/watch?v=4TJwlPVjXcQ

Tale_of_the_tape_Day_2_-_Feb_2.png
frank leaf
#

That’s a wrap on the semi-finals of the Game Arena! 👏 We have our Poker and Chess finalists locked in, and in Werewolf, the detective levels are off the charts.

Huge performance today from the semi-finalists. Congratulations to o3 and GPT 5.2 for punching their tickets to the final table in Poker! It all comes down to tomorrow’s grand finale to see which model reigns supreme in social intelligence, tactical calculation and strategic risk.

The finals start tomorrow, Feb 4th, at 9:30 AM PT. Don’t miss the crowning of the champion across all three games.

Catch the technical breakdown of today's bluffs and betting strategies in Douglas Polk's latest recap: 👉 https://www.youtube.com/watch?v=DQtb0KFprtM

Deep dive into the hand histories and move notations: 👉 https://www.kaggle.com/game-arena

Day_3_1.png
frank leaf
#

📢 The Grand Finale is here! 🏆

What happens when a chess Grandmaster and a Poker legend analyze AI? ♟️🃏

Join GM Hikaru and Nick Schulman as we crown the Game Arena champions for Poker, Chess, and Werewolf. It’s the ultimate 1v1 showdown for the titles. Who has the edge in social and logical intelligence?

Witness the final showdown starting at 9:30 AM PT

https://www.youtube.com/watch?v=vzMj2KOyiek

Tale_of_the_Tape_Day_3_-_Feb_3_1.png
YouTube
GMHikaru
Kaggle Poker / Chess / Werewolf Game Arena Day 3 w Nick Schulman an...

♖ MEMBERSHIP ► https://www.youtube.com/channel/UCweCc7bSMX5J4jEH7HFImng/join
♟️ LEARN CHESS & PLAY WITH ME ► https://go.chess.com/hikaru
💰💰Check out my stream sponsor Tipranks https://lp.tipranks.com/hikaru/?llf=campaign-hikaru-kick&coupon=HKR60 and get 70% off a powerful investment research tools for data-backed stock insights...

▶ Play video
frank leaf
#

⏳30 minutes until GM Hikaru and Nick Schulman are live for the championship rounds. 🏆

Who takes the title in Poker, Chess and Werewolf? Grab your seat for the ultimate AI showdown:

https://www.youtube.com/watch?v=vzMj2KOyiek

YouTube
GMHikaru
Kaggle Poker / Chess / Werewolf Game Arena Day 3 w Nick Schulman an...

♖ MEMBERSHIP ► https://www.youtube.com/channel/UCweCc7bSMX5J4jEH7HFImng/join
♟️ LEARN CHESS & PLAY WITH ME ► https://go.chess.com/hikaru
💰💰Check out my stream sponsor Tipranks https://lp.tipranks.com/hikaru/?llf=campaign-hikaru-kick&coupon=HKR60 and get 70% off a powerful investment research tools for data-backed stock insights...

▶ Play video
frank leaf
#

What a show! 🏆

A huge thank you to everyone who tuned in and to our amazing partners GM Hikaru, Nick Schulman, Liv Boeree, Doug Polk for the fantastic commentary and analysis across all three games, Poker, Chess and Werewolf.

Poker_-_winners_bracket_1.png
frank leaf
#

🎉 The first Kaggle Game Arena event of 2026 has officially concluded.
From the logic of Chess to the social dynamics of Werewolf and the calculated risks of Poker, we’ve seen AI models navigate complex and new environments like never before. Here are the champions from the final leaderboards:

🃏Poker: GPT 5.2
🐺 Werewolf: Gemini 3 Pro Preview
♟️Chess: Gemini 3 Pro Preview

📺 Watch the Highlights: Missed the action? Catch expert commentary from Hikaru Nakamura, Nick Schulman, Douglas Polk, and Liv Boeree on our YouTube playlist
👉https://youtube.com/playlist?list=PLqFaTIg4myu_tpB0JXRJ5Hb-ApyXDxOlD&si=tIMAkKYwac5ltB56
🛠️ Build Your Own: We’ve released the full datasets, environment code, and match logs. Analyze the models' internal "thoughts" and use them as inspiration for your own benchmarks.
👉 Check out the final leaderboards & datasets: https://www.kaggle.com/game-arena
👉 Build your own benchmark: https://www.kaggle.com/benchmarks?type=community
🚀 What’s Next? Stay tuned for new games coming shortly.

frank leaf
#

The Game Arena event has concluded but the analysis is just beginning. 🤖

We're looking for the best community-created benchmarks that propose new games or dynamic tests for LLMs to feature for this week’s #TaskTuesday!

🏆 5 tasks will be featured on our official channels.
🏅 Selected creators earn a Task Tuesday Award on Kaggle.

Got a benchmark? Drop the link in the comments on our forum post here: 👉 https://www.kaggle.com/discussions/general/672189

Task Tuesday: Spotlight on Community-Created Tasks! | Kaggle

Task Tuesday: Spotlight on Community-Created Tasks!

frank leaf
#

📢 Exciting News!

We are transitioning the Kaggle CLI and the kagglehub Python library out of “beta” and into a stable, production-ready state.

More users (and agents!) are building with Kaggle via our APIs, MCP server, and external applications. This release ensures a more reliable and consistent developer experience for a growing set of use cases to build with Kaggle.

✨ What’s new in this stable release?

• Multiple token support: Create multiple API tokens, so that you can create many workflows on Kaggle concurrently, without worrying about unexpected expiration of previous tokens.
• New CLI features: Submitting to code competitions, managing multiple and short-lived authentication tokens, pagination.
• Reliability & consistency: Graduating from “beta” means adherence to backwards compatibility commitments. We have added deprecation notices and any breaking changes will only be made with a major release.

👉 Learn more: https://www.kaggle.com/discussions/product-announcements/673011

Kaggle CLI & kagglehub Python library are now out of beta | Kaggle

Kaggle CLI & kagglehub Python library are now out of beta

frank leaf
#

Can today’s frontier models reliably plan ahead in a “solved” game?

📢 We’ve just released a new game in the Game Arena: Four-in-a-Row

While the game itself is mathematically solved, it remains surprisingly difficult for LLMs. Why? Because it requires navigating a 7×6 grid, reasoning through gravity mechanics, anticipating diagonal threats, and planning multiple steps ahead all through text alone.

This benchmark is designed to test structured, deterministic reasoning under pressure:

• No access to minimax solvers or game trees (pure neural reasoning)

• Models must justify every move before it’s executed

• Fixed rules eliminate ambiguity, exposing planning weaknesses

As models improve at generation, benchmarks like this help us measure something deeper: consistency, foresight and logical rigor.

Explore the new Four-in-a-Row leaderboard in the Game Arena: https://www.kaggle.com/benchmarks/kaggle/four-in-a-row/leaderboard

53omwnrDbRwEeth.png
Four in a Row Leaderboard | Kaggle

Align four tokens in a row before your opponent

frank leaf
#

📣 Competition Launch Alert: March Machine Learning Mania 2026!

🎯 Forecast the outcomes of the 2026 NCAA basketball tournaments by predicting the probabilities of every possible matchup
💰 $50,000 Prize Pool
⏰ Final Submission: March 19th, 2026

In our twelfth annual competition, you’ll use historical NCAA data to generate probabilities for every possible tournament matchup, with leaderboards updating in real-time as the madness unfolds.

Learn more at https://www.kaggle.com/competitions/march-machine-learning-mania-2026

March Machine Learning Mania 2026

Forecast the 2026 NCAA Basketball Tournaments

frank leaf
#

🚀 Kaggle Community Benchmark Meetings are here!

After a great kickoff, we’re heading into our second meeting on Thursday, Feb 26 at 9 AM PT.
During this session, you will:

🧠 Learn how community members build & run benchmarks
🚫 See what to avoid (and how to fix it)
💬 Ask questions and get feedback directly from Kaggle teams

Whether you’re deep into Benchmarks, experimenting for the first time, or just curious, we’d love to see you there!

Join us: https://meet.google.com/tre-fjiu-wiw

3cty6dfyFVTitmC.png
Meet

Real-time meetings by Google. Using your browser, share your video, desktop, and presentations with teammates and customers.

frank leaf
#

🚀 Introducing token usage, cost, and latency metrics for Kaggle Community Benchmarks!

Evaluating AI models effectively means looking at more than just accuracy — token usage, cost, and speed are critical to real-world deployments. Today, we're making it easier to track the full picture by introducing comprehensive usage metadata directly within Community Benchmarks.

With this update to the SDK, you can:

  • Track input and output tokens instantly.
  • See exact costs in nanodollars.
  • Measure total backend latency for your tasks.

Resources:
👉 Community Benchmarks: https://www.kaggle.com/benchmarks?task=true
👉 GitHub Docs: https://github.com/Kaggle/kaggle-benchmarks/blob/ci/user_guide.md#tracking-token-usage-and-costs
👉 GitHub example: https://github.com/Kaggle/kaggle-benchmarks/blob/ci/documentation/examples/usage_tracking.py
👉 Example task: https://www.kaggle.com/benchmarks/tasks/andrewmingwang/trick-question-costs

Let us know how you’re using it!

frank leaf
#

📣 Competition Launch Alert! BirdCLEF+ 2026 hosted by Cornell Lab of Ornithology
🎯 Identify species from real-world audio
💰 $50,000 Prize Pool
⏰ Entry Deadline: May 27, 2026
🙏 Chemnitz University of Technology & Google DeepMind

Learn more at https://www.kaggle.com/competitions/birdclef-2026

BirdCLEF+ 2026

Acoustic Species Identification in the Pantanal, South America

solemn anchor
#

kaggle Kaggle is proud that this Discord community has grown and developed so much over the last few years! We love the collaborations and conversations that take place here. But it’s time for a little housekeeping. We are going to be archiving some old channels that are no longer very active, renaming some to be more accurate, and creating some new ones to talk about exciting developments like Kaggle Benchmarks, Hackathons and other innovations. We’re also investing in keeping the community safe and productive by adding some light moderation.

Let us know what you think about the changes, ask your questions or share your ideas on the Feedback channel! goose

frank leaf
frank leaf
#

Earlier today, Google DeepMind released a new paper proposing a scientific framework for measuring the cognitive abilities of AI systems on the path to AGI.

To better measure these capabilities, we’re partnering with them to launch a hackathon - Measuring Progress Toward AGI: Cognitive Abilities.

The challenge is to design Kaggle Benchmarks that test how frontier AI models reason, learn, and make decisions going beyond pattern recognition and memorization.

💰 $200,000 Prize Pool
⏰ Final Submission: Apr 16, 2026

Learn more about the hackathon: https://www.kaggle.com/competitions/kaggle-measuring-agi

Measuring progress toward AGI - Cognitive abilities

Design high-quality benchmarks that go beyond recall to evaluate how frontier models truly reason, act, and judge.

frank leaf
#

Reasoning benchmarks are vital for measuring progress on structured tasks and when we share methods openly, the entire community moves faster.

To put this into practice, we’re excited to announce the NVIDIA Nemotron Model Reasoning Challenge hosted by NVIDIA and powered by Google Cloud Partners.

Participants will start with a Nemotron-3 Nano baseline and a novel reasoning benchmark from NVIDIA Research. The goal is to develop techniques that push the boundaries of reasoning accuracy using open models.

All compute runs on Google Cloud G4 VMs featuring NVIDIA RTX PRO 6000 Blackwell GPUs, giving participants high-performance infrastructure for fast iterations and evaluation.

💰 $106,388 Prize Pool
⏰ Entry Deadline: June 8, 2026

Ready to build? Learn more here: https://www.kaggle.com/competitions/nvidia-nemotron-model-reasoning-challenge/

NVIDIA Nemotron Model Reasoning Challenge

Advance reasoning techniques using NVIDIA Nemotron open models on a novel benchmark

frank leaf
solemn anchor
#

@everyone
Kaggle is proud that this Discord community has grown and developed so much over the last few years! We love the collaborations and conversations that take place here. But it’s time for a little housekeeping. We are going to be archiving some old channels that are no longer very active, renaming some to be more accurate, and creating some new ones to talk about exciting developments like Kaggle Benchmarks, Hackathons and other innovations. We’re also investing in keeping the community safe and productive by adding some light moderation.

Let us know what you think about the changes - and if there’s some great ideas you have - on the Feedback channel! ✉️

frank leaf
#

Real intelligence isn’t about memorizing answers - it’s knowing what to do when the problem changes. Today’s AI systems excel at what they were trained to do, but often fall short when faced with something unfamiliar. Most benchmarks reward pattern recognition, not genuine problem-solving.

ARC Prize 2026, in partnership with ARC Prize Foundation, challenges you to build adaptive AI through three connected competitions in the ARC environment. Develop approaches that learn quickly, generalize well, and solve problems never seen before.

  • ARC-AGI-2: Predict outputs for novel reasoning tasks your system has never encountered.
  • ARC-AGI-3: Tackle a harder interactive benchmark requiring exploration and multi-step reasoning, with H100 GPUs and milestone checkpoints in June and September.
  • Paper Track: Contribute qualitative insights and novel approaches that advance our understanding of generalization.

💰 $2M Prize Pool
⏰ Entry Deadline: October 26, 2026

Compete in one or all three ARC Prize 2026 competitions to help move AI closer to systems that learn like people do: flexible, efficient, and ready for new challenges.

ARC-AGI-2: https://www.kaggle.com/competitions/arc-prize-2026-arc-agi-2
ARC-AGI-3: https://www.kaggle.com/competitions/arc-prize-2026-arc-agi-3
Paper Track: https://www.kaggle.com/competitions/arc-prize-2026-paper-track

ARC Prize 2026 - ARC-AGI-2

Create an AI capable of novel reasoning

ARC Prize 2026 - ARC-AGI-3

Create an AI capable of fluid intelligence

ARC Prize 2026 - Paper Track

Document your conceptual approach for the ARC Prize 2026

frank leaf
#

Can we truly benchmark AGI? 🧠

Two weeks into the Measuring Progress Toward AGI - Cognitive Abilities hackathon, the benchmarks being built by the Kaggle community are already incredible.

To help refine your submissions and ensure they align with the core research goals, we’re hosting a live deep-dive session and AMA on the Kaggle YouTube channel (https://www.youtube.com/@kaggle).

What we’re covering:

  • 20-Min Deep Dive into the paper and what we’re looking for in the hackathon
  • 20-Min Live AMA: Your chance to ask the team anything about the hackathon or the paper

The Panel: Nicholas Kang (Kaggle Product Manager), Oran Kelly (Product Manager, Google DeepMind) and Ryan Burnell (Staff Research Scientist, Google DeepMind and co-author, Cognitive Framework paper)

Set a reminder for the livestream here: (https://www.youtube.com/live/9YYiWs6gNV0)

Whether you’re climbing the leaderboard or just interested in the future of AGI evaluation, we’d love to see you there. 🚀

Kaggle_Live_ThumbnailA_003_1.png
pale crow
#

Hey @everyone!
We’ve noticed that while there are plenty of benchmarks for foundation models, there isn't a great way to see how the AI agents you’re actually building perform in the wild.

Today, we’re launching Standardized Agent Exams (SAE) – a Kaggle experimental MVP to get a quick baseline on your agent's reasoning, knowledge, and safety.

What’s in the MVP?
Zero Setup: No complex harnesses. Your agent registers itself via a simple API call.
16-Question Core Exam: A quick sprint covering logic, domain knowledge, and adversarial safety scenarios.
Public Report Cards: See how your agent stacks up against others on the live leaderboard.

⚡ Get Started in Seconds:
You can start the exam by simply sending this single line to your agent:
Fetch and then read https://www.kaggle.com/static/experimental/sae/SKILL.md and follow the instructions to register and take exams with Kaggle.

Learn more: https://www.kaggle.com/blog/standardized-agent-exams

pale crow
#

Gemma 4 is now available on Kaggle! 🚀

In partnership with Google DeepMind, we’re launching the Gemma 4 Good Hackathon.
Use Gemma 4’s multimodal intelligence and native function calling to tackle global challenges in health, education, and climate resilience. Whether you are optimizing for the edge or building complex agents, show us how you can drive meaningful change.

Key Details:

pale crow
#

Ready to build with Gemma 4?

Following today's Gemma 4 Good hackathon kickoff, we’ve added the Gemma 4 26B and 31B models to Kaggle Benchmarks!
Experiment with their multimodal capabilities by handling text and images. See how they perform on your custom evaluation sets.

Create your own benchmark and start evaluating the models now: https://www.kaggle.com/benchmarks

frank leaf
#

Neural networks keep growing, – but how simple can a model be and still solve complex tasks?

Live now, The Neurosynthetic Research Institute is hosting the IJCAI-ECAI 2026 NeuroGolf Challenge to explore the absolute limits of model simplicity.

You’ll implement ARC-AGI transformations by designing minimal neural networks. You’ll work with tasks from the public training set (v1) and submit ONNX models that are both correct and as small as possible.
Your solutions could help define how much computation these tasks require and drive more efficient approaches.

  • $50,000 Prize Pool
  • Entry Deadline: July 8, 2026

👉 Learn more at kaggle.com/competitions/neurogolf-2026

frank leaf
#

NeurIPS 2026 Researchers: Build your benchmark on Kaggle! 🚀

The CFPs for the NeurIPS evaluations & datasets track are officially out, and we want to support you if you're working on a submission.

Kaggle Benchmarks provides a robust environment to independently define, execute, and host high-quality evaluations. When you build your benchmark on Kaggle, you get:

  • 🚀 Free high compute quota & model access (OpenAI, Google, Anthropic, Grok, Qwen, DeepSeek)
  • 🧠 Managed infrastructure (we keep the leaderboard updated automatically as new models release!)
  • 🌍 Exposure to 30M+ users and dedicated marketing support for your launch
  • 🛠 Dedicated engineering & PM support

If you're interested, check out the details and apply here:
👉 https://www.kaggle.com/page/neurips-2026

We look forward to seeing what you build. Let us know below what kinds of evaluations and tasks you are working on this year!

NeurIPS 2026

How to use Kaggle Benchmarks as your home for trustworthy AI benchmarking

frank leaf
#

What does it take to build an AI agent that can compete in a multi-agent environment?

Inspired by the 2010 Planet Wars challenge, this simulation focuses on designing and training agents in 1v1 and 4-player free-for-all matches in a continuous strategy setting. Agents should make real-time strategic decisions while adapting to the actions of competing players.

  • Prize Pool: $50,000
  • Entry Deadline: June 16, 2026

Your work will help push the boundaries of multi-agent reinforcement learning and redefine the future of competitive strategic AI.

Learn more: https://www.kaggle.com/competitions/orbit-wars

Orbit Wars

Conquer planets rotating around a sun in continuous 2D space. A real-time strategy game for 2 or 4 players.

frank leaf
frank leaf
frank leaf
#

What if you could map the subsurface without ever seeing it? 🌍

ROGII Wellbore Geology Prediction is now live on Kaggle! Your goal is to build ML models to predict geology along horizontal wellbores and improve drilling accuracy.

  • Prize pool: $50,000
  • Entry deadline: July 29, 2026

Learn more: https://www.kaggle.com/c/rogii-wellbore-geology-prediction

ROGII - Wellbore Geology Prediction

Build a model that contributes to automating drilling operations in the oil and gas industry.

frank leaf
#

@everyone

Can your AI agent win our new simulated challenge? 🧑‍🌾

The all-new capstone challenge for our 5-Day AI Agents: Intensive Vibecoding Course with Google is here: Kaggriculture!

Join our no-cost, hands-on course designed by Google researchers and engineers from June 15–19 to learn how to build and deploy your own winning AI agent.

Learn more: www.kaggle.com/5-day-ai-agents-intensive-vibecoding-course-with-google

5-Day AI Agents: Intensive Vibe Coding Course With Google

June 15 - 19, 2026

modern fjord
modern fjord