Hacker News new | ask | show | jobs
Anthropic NLAs translate LLM activations to human-readable text for safety (presciente.com)
1 points by sebastianperezr 46 days ago