Research tests whether AI models know their own thoughts

Researchers inject thoughts into large language models to probe their capacity for introspection.