Miles Redcliffe; Matthew Blackwood. Improving Interpretability in Large Language Models through Layer Wise Relevance Propagation for Identifying Latent Reasoning Biases. IJAIR 2026, 1.