@article{flashheadefficientdropinreplacementforth, title = {FlashHead: Efficient Drop-In Replacement for the Classification Head in Language Model Inference}, author = {Wilhelm Tranheden and Shahnawaz Ahmed and Devdatt Dubhashi and Jonna Matthiesen and Hannes von Essen}, year = {2026}, eprint = {2603.14591}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2603.14591}, }