/
#2
Anthropic
AI safety company building the Claude LLM family (Haiku, Sonnet, Opus). Training focuses on Constitutional AI and RLHF for alignment. Research spans mechanistic interpretability via dictionary learning, scalable oversight, and frontier red-teaming. Models support extended thinking, tool use, and agentic workflows; available via API, AWS Bedrock, and Google Vertex AI.
Subtopics
POLICYDEFENSERED TEAMING
Links
LAST 30 DAYS
