logoalt Hacker News

7moritz7yesterday at 8:28 PM0 repliesview on HN

These are sota models, not open source 7b parameter ones. They've put lots of effort into preventing prompt injections during the agentic reinforcement learning