Tag: agentic AI safety
Browse our exclusive articles!
Controlling LLM Sycophancy by Verbalizing Assumptions
Explore how verbalizing assumptions in LLMs helps explain and control sycophantic behavior, improving AI safety and response objectivity.
OpenAI Safety Fellowship: AI Research Grants 2024
Apply for the OpenAI Safety Fellowship to get funding, mentorship, and support for AI safety and alignment research. Join now to advance ethical AI.
Token-Space Attacks on Reward Models in RLHF
Discover how token-space attacks exploit reward models in RLHF, revealing vulnerabilities beyond semantic manipulation and impacting AI safety.
Limits of Reinforcement Learning Alignment in AI Safety
Explore the generalization limits of reinforcement learning alignment and its impact on AI safety in large language models with compound jailbreaks analysi...
Impact of Safety Unalignment on Large Language Models
Explore how safety unalignment affects large language models, highlighting risks, performance changes, and mitigation strategies for safer AI deployment.