Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
Adversarial GLUEadversarialglue.github.io · Low evidenceAdvGLUE is an adversarial robustness evaluation benchmark that tests language models across five natural language understanding tasks. It covers diverse adversarial attacks including word-level, sentence-level, and human-written examples to comprehensively assess model vulnerabilities.
Parse Score