@article{howdoinstructionsshapespeechcrossattenti,
  title = {How Do Instructions Shape Speech? Cross-Attention Attribution for Style-Captioned Text-to-Speech},
  author = {Nityanand Mathur and Hamees Sayed and Wasim Madha and Apoorv Singh and Sameer Khurana and Akshat Mandloi and Sudarshan Kamath},
  year = {2026},
  eprint = {2606.20532},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.20532},
}