Andrey Kurenkov

Scores of R1, Flash-thinking, Claude 4.7, QwQ, o1-pro, o3-mini on USAMO 2025. Scores less than 5% of max score.

Tests on USAMO immediately after problems were posted yield surprisingly bad model performance. Suggests there's much more training on test than expected.
arxiv.org/abs/2503.219...

Merry Christmas to all the cats enjoying their empty boxes this year! (And to everyone else too!)🎄🐾

Merry Christmas.
Cartoon by the wonderful Liz Climo

but the Greatest Gift of All is nine months of no Christmas music

A weird thing with AI is that so much has gotten so mind blowing so quick we are kind of immune to it by now...

I mean, that AI can make weird but realistic videos and funny songs from a prompt like this is crazy!!!

YouTube video by Last Week in AI Current of Dreams - AI Music Video

My favorite AI creation so far
youtu.be/QIxwtzKAXTg?...

thinking of calling this "The Illusion Illusion"

(more examples below)

you can post jokes on here if you want. it's legal

Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions Currently OpenAI o1 has sparked a surge of interest in the study of large reasoning models (LRM). Building on this momentum, Marco-o1 not only focuses on disciplines with standard answers, such as mat...

Alibaba has their own version on GPT-o1. This might be the best description of “o1-type”systems so far arxiv.org/abs/2411.14405

Some machine learners were once children. Here’s where you can find them:

go.bsky.app/F6mM37U

Yeah, thanks!

Hi #AI folks on Bluesky!

Been a looooong while since I stopped being active on Twitter/Mastodon, but hey let's give this a try and see if it sticks.

Here are some hashtags in case that does something: #AIResearch #DeepLearning #MachineLearning #Robotics #LLM