Jerry Tworek

“Co-author of the GSM8K dataset paper on training verifiers to solve math word problems.”

“Commented on limited AI-attributed progress in nanoGPT speedruns.”

“Former OpenAI researcher who launched Core Automation, an AI lab aiming to build the most automated AI lab in the world, developing new learning algorithms beyond pre-training and reinforcement learning.”

“Former OpenAI research VP who founded Core Automation, a new AI lab focused on developing AI systems to automate AI development itself.”

Google Gemini Deep Think and OpenAI's experimental reasoning model both achieved International Math Olympiad Gold medal performance by solving 5 of 6 problems within human time limits using general-purpose reinforcement learning and test-time compute scaling.

“Commented that OpenAI did very little IMO-specific work and focused on training general models in natural language.”