Doesn’t mean LLMs aren’t incredibly powerful as noted in the thread. Just demonstrates the limits of how simulating plausible answers breaks down when you need chains of exact ones.

Posters: eshear. eshear

2 likes recorded by the source corpus.