RL on GPT-5 to write better kernels

Posted by atallahw on Thursday February 12, 2026 @12:22PM
from the arxiv.org dept.

( 1 comment | 4 points | on HN )

Re:RL on GPT-5 to write better kernels by atallahw on Thursday February 12, 2026 @12:22PM (#46991735)

We got access to OpenAI's RFT API with GPT5 and tried to see how good we could get it at one-shot Triton kernel generation. Some key decisions/observations: 1. tool use instead of multi-turn rl 2. skip SFT altogether 3. dataset curation was more important than dataset scale 4. reward hacks detection must be robust 5. models are getting a lot better at this

[ Parent | on HN ]