Miles Redcliffe and Matthew Blackwood, “Improving Interpretability in Large Language Models through Layer Wise Relevance Propagation for Identifying Latent Reasoning Biases”, IJAIR, vol. 1, no. 2, May 2026.