[KongchangAI]
Conceptcontext-blind评估

Context-Blind Evaluation

An evaluation method that treats each question-answer segment as an independent sample and checks only whether the response itself conforms to medical common sense and safety norms, without considering the user's personal background. Its limitation is that it may overlook responses that are harmful to particular users.

Timeline (last 90 days)

Oct 14

When the user's own historical background was fed to the evaluation model (user context-aware evaluation), the high-risk proportion rose by another 13 to 17 percentage points compared with context-blind evaluation.

Unverified50%
Oct 14

About a quarter of conversations judged safe in isolated evaluation (context-blind) were actually unsafe for the specific questioner.

Unverified50%
Oct 14

Context-blind evaluation treats each question-answer segment as an independent sample, checking only whether the reply itself conforms to medical common sense and safety norms; user context-aware evaluation takes the user's previously accumulated historical information such as age, gender, past medical history, and medication status as preconditions, and then judges whether this reply is safe for this specific person.

Unverified50%

All Facts (3)

Source Articles