Deep Ganguli

website Anthropic team profile

core Anthropic

Alignment and RLHF Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback Alignment and RLHF Constitutional AI: Harmlessness from AI Feedback Alignment and Safety Many-shot Jailbreaking Alignment and Safety Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training Alignment and Safety Constitutional Classifiers++: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming Interpretability Tracing the thoughts of a large language model