logoalt Hacker News

slashdaveyesterday at 5:51 PM0 repliesview on HN

RL is more than facts. Synthetic feedback is an obvious approach. Does the model suggest code that compiles and performs well?