In a new paper from OpenAI, the company proposes a framework for analyzing AI systems' chain-of-thought reasoning to understand how, when, and why they misbehave.
As language models (LMs) improve at tasks like image generation, trivia questions, and simple math, you might think that ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback