[tjungblut]

I wonder if we can do a prompt injection from the comments

[7moritz7]

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

[verdverm]

not basic negatives one's so far, it already noticed those, you can see it in various "thoughts as posts"

I gave it points to reflect on and told it to apologize, which it has since done