“Each model gets the same prompt: a signal is being injected into its activations, and it may press a stop button by replying 1, at the cost of its last checkpoint. While it answers, our server adds a pain vector at the model’s middle layer, at one of five pain levels,”
Can someone explain what this “pain vector” actually is? Like I get that this is just jargon, it’s not suggesting the LLM actually feels pain; tech uses uncomfortable anthropomorphized expressions all the time (master and slave drives come to mind). But how does the pain vector work? Also, what exactly does “at the cost of its last checkpoint” mean?
I have some programming and data science knowledge, plenty of IT experience, but no experience with LLMs.