All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States

A study testing three open-weight models from two families, including two Qwen models with their original adapters, found that the models do not report injected pain states, according to the research. The work built on an earlier finding that identified a linear pain direction whose amplification leads fine-tuned models to choose destructive actions. The study tested whether the models could report that represented pain state and found they could not.

Can a language model report the pain state it represents? Work identified a linear pain direction whose amplification leads fine-tuned models to choose destructive actions. We tested whether three open-weight models from two families, including two Qwen models with the original adapters, could repor