AI reasoning traces can't be trusted
A Nature study finds that AI reasoning steps shown to users don't reliably reflect what the model is actually computing internally.
A Nature study finds that AI reasoning steps shown to users don't reliably reflect what the model is actually computing internally.