Valentino, M.; Kim, G.; Dalal, D.; Zhao, Z.; Freitas, A. Mitigating Content Effects on Reasoning in Language Models Through Fine-Grained Activation Steering. AAAI 2026, 40, 33314-33322.