CALM : A Multi-task Benchmark for Comprehensive Assessment of Language Model Bias
This work addresses the need for reliable bias assessment in language models for AI ethics and fairness, though it is incremental as it builds on existing bias measurement frameworks.
The authors tackled the problem of robustly measuring sociodemographic bias in language models by introducing CALM, a multi-task benchmark that integrates 16 datasets across three NLP tasks, resulting in more stable bias scores with reduced sensitivity to template perturbations compared to prior methods.
As language models (LMs) become increasingly powerful and widely used, it is important to quantify them for sociodemographic bias with potential for harm. Prior measures of bias are sensitive to perturbations in the templates designed to compare performance across social groups, due to factors such as low diversity or limited number of templates. Also, most previous work considers only one NLP task. We introduce Comprehensive Assessment of Language Models (CALM) for robust measurement of two types of universally relevant sociodemographic bias, gender and race. CALM integrates sixteen datasets for question-answering, sentiment analysis and natural language inference. Examples from each dataset are filtered to produce 224 templates with high diversity (e.g., length, vocabulary). We assemble 50 highly frequent person names for each of seven distinct demographic groups to generate 78,400 prompts covering the three NLP tasks. Our empirical evaluation shows that CALM bias scores are more robust and far less sensitive than previous bias measurements to perturbations in the templates, such as synonym substitution, or to random subset selection of templates. We apply CALM to 20 large language models, and find that for 2 language model series, larger parameter models tend to be more biased than smaller ones. The T0 series is the least biased model families, of the 20 LLMs investigated here. The code is available at https://github.com/vipulgupta1011/CALM.