Cross-Care: Assessing the Healthcare Implications of Pre-training Data on Language Model Bias

arXiv:2405.05506v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly essential in processing natural languages, yet their application is frequently compromised by biases and inaccuracies originating in their training data. In this study, we introduce Cross-Care, the first benchmark framework dedicated to assessing biases and real world knowledge in LLMs, specifically focusing on the representation of disease prevalence across diverse demographic groups. We systematically evaluate how demographic biases embedded in pre-training corpora like $ThePile$ influence the outputs of LLMs. …

abstract application arxiv benchmark bias biases cs.cl data framework healthcare language language model language models languages large language large language models llms model bias natural pre-training processing study training training data type world