Press Space to continue
Weak AI models can supervise training of stronger ones
Press Space to continue
New Anthropic Fellows research: developing an Automated Alignment Researcher. We ran an experiment to learn whether Claude Opus 4.6 could accelerate research on a key alignment problem: using a weak AI model to supervise the training of a stronger one. https://anthropic.com/research/automated…
Automated Alignment Researchers: Using large language models to scale scalable oversight