Ars Technica on Nostr: Why do LLMs make stuff up? New research peers under the hood. Claude's faulty "known ...
Why do LLMs make stuff up? New research peers under the hood.
Claude's faulty "known entity" neurons sometime ovverride its "don't answer" circuitry.
https://arstechnica.com/ai/2025/03/why-do-llms-make-stuff-up-new-research-peers-under-the-hood/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social
Claude's faulty "known entity" neurons sometime ovverride its "don't answer" circuitry.
https://arstechnica.com/ai/2025/03/why-do-llms-make-stuff-up-new-research-peers-under-the-hood/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social