logoalt Hacker News

irthomasthomastoday at 12:19 AM0 repliesview on HN

multi-model arbitration, synthesis, parallel reasoning etc. Judging large models with small models is quite effective.