XAI-CRED: Explainable Artificial Intelligence and Credibility in Asylum Decision-making

LLMs as annotators of credibility assessment in Danish asylum decisions: evaluating classification performance and errors beyond aggregated metrics