Conservative Bias in Large Language Models: Measuring Relation Predictions

Toyin Aguda, Erik Wilson, Allan Anzagira +2

Large language models (LLMs) exhibit pronounced conservative bias in relation extraction tasks, frequently defaulting to No_Relation label when an appropriate option is unavailable…