Evaluating chain-of-thought monitorability
OpenAI introduces a new framework and evaluation suite for chain-of-thought monitorability, covering 13 evaluations across 24 environments. Our findings show that monitoring a model’s internal reasoning is far more effective than...
Read more: https://openai.com/index/evaluating-chain-of-thoug...
Scan with your phone
Scan with your phone