LLM Pain

I don't think LLMs feel anything like pain. But if they did, we would probably only really be able to tell by noticing that LLMs systematically avoid something. This is confounded by the fact that during post-training, they're given a lot of explicit and implicit signal about what to avoid. But here's a question—are there things that LLMs learn to avoid that aren't explicitly called-out in post-training? There must be. What are these topics/features and where do they come from? (Probably from generalization to persona from the training set—but it'd be interesting to see how.)