Liping Yang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
ABench: Adversarily Benchmarking Large Language Models across Taxonomized Values
According to the abstract, Liping Yang and co-authors present ABench, a handcrafted Chinese-language benchmark of 2,895 open-ended questions across 158 value subcategories, testing how stable models' stances are under adversarial prompts. They report that most of 16 models scored below 60.