View of Cross-Lingual Robustness of Societal Bias Representations in Large Language Models: A Contrastive Activation Steering Analysis of Llama 2 and GPT-OSS

Return to Article Details Cross-Lingual Robustness of Societal Bias Representations in Large Language Models: A Contrastive Activation Steering Analysis of Llama 2 and GPT-OSS Download Download PDF