Anthropic NLAs translate LLM activations to human-readable text for safety

Posted by sebastianperezr 4 hours ago

1 points | 0 comments