Counterfactual audit of ten open-source LLMs for pediatric ED triage examines how injected demographic and socioeconomic variables affect acuity predictions. The study spans model families, sizes, medical-domain…

#OpenSourceAI #ClinicalAI #LLMBias #AIResearch
https://arxiv.org/abs/2610.01963

Studying political bias in AI writing assistants, researchers tested six models with 148,302 prompts on 107 Swedish policy issues ahead of the 2026 election. The work offers a structured way to probe default LLM stances…

#GenAI #LLMBias #Prompting #AIResearch
https://arxiv.org/abs/2609.15207