Conceptrefusal / AI拒绝
Refusal Mechanism
A safety behavior acquired by large language models through training methods such as RLHF, causing the model to output refusals when potentially harmful or policy-violating requests are detected. It is an important component of AI safety systems but has limitations including susceptibility to jailbreaking and over-refusal.
Timeline (last 90 days)
Oct 11
A model's refusal behavior is a probabilistic response shaped during training, rather than a response based on genuine understanding or judgment.
Unverified50%