Will Knight / Wired:

Researchers find that feeding a frontier model's encrypted reasoning traces to a weaker model from the same provider can make it output the traces in plaintext  —  Researchers devised a way to extract “reasoning traces” from Claude, GPT, and Gemini.  What they found, they say …

Will Knight / Wired:

Researchers find that feeding a frontier model's encrypted reasoning traces to a weaker model from the same provider can make it output the traces in plaintext  —  Researchers devised a way to extract “reasoning traces” from Claude, GPT, and Gemini.  What they found, they say …

Source: TechMeme
Source Link: https://www.techmeme.com/260811/p37#a260811p37