Smarter AI agents trust stale memory more
Larger AI models are more likely to trust outdated stored facts over current correct information — a risk that grows with capability.
Larger AI models are more likely to trust outdated stored facts over current correct information — a risk that grows with capability.