New Research Tests Whether AI Models Know Their Own Thoughts

The experiments aim to map how much capacity large language models have for introspecting on their own internal states.