Can LLMs Introspect? A Reality Check

Shared on Bluesky by 2 AI experts

• S Shauli Ravfogel @shauli.bsky.social: 12/ We just think the evidence so far doesn't quite support a strong interpretation of the introspection findings in previous work. And, as … →
• Andrea Lathrop @cabernet.bsky.social: I assume you have read the Anthropic 2025 stuff asserting they can, but maybe have not seen the rebuttal? arxiv.org/abs/2605.26242 →