A framework for dynamically generating context-aware bias and stereotyping test prompts for LLMs, with no empirical validation.
Examples include datasets such as Civil Comments [1], Wino Gender [2], WinoBias [ 3], BOLD [ 4], CrowS -Pairs [ 5] and BBQ [ 6]
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
ASCenD-BDS: Adaptable, Stochastic and Context-aware framework for Detection of Bias, Discrimination and Stereotyping
A framework for dynamically generating context-aware bias and stereotyping test prompts for LLMs, with no empirical validation.