%Aigaion2 BibTeX export from Idiap Publications
%Thursday 04 December 2025 06:19:21 PM

@INPROCEEDINGS{Kumar_IEEEASRU2025_2025,
                      author = {Kumar, Shashi and Madikeri, Srikanth and Villatoro-Tello, Esa{\'{u}} and Burdisso, Sergio and Rangappa, Pradeep and Carofilis, Andr{\'{e}}s and Motlicek, Petr and S, Karthik Pandia D and Venkatesan, Shankar and Hacioğlu, Kadri and Stolcke, Andreas},
                    keywords = {language identification, multitask training, named entity recognition, Speaker change detection, speech recognition, XLSR-Transducer},
                    projects = {UNIPHORE, ELOQUENCE},
         mainresearchprogram = {Human-AI Teaming},
  additionalresearchprograms = {AI for Everyone},
                       title = {TokenVerse++: Towards Flexible Multitask Learning with Dynamic Task Activation},
                   booktitle = {2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)},
                        year = {2025},
                   publisher = {IEEE},
                    abstract = {Token-based multitasking frameworks like TokenVerse require all training utterances to have labels for all tasks, hindering their ability to leverage partially annotated datasets and scale effectively. We propose TokenVerse++, which introduces learnable vectors in the acoustic embedding space of the XLSR-Transducer ASR model for dynamic task activation. This core mechanism enables training with utterances labeled for only a subset of tasks, a key advantage over TokenVerse. We demonstrate this by successfully integrating a dataset with partial labels, specifically for ASR and an additional task, language identification, improving overall performance. TokenVerse++ achieves results on par with or exceeding TokenVerse across multiple tasks, establishing it as a more practical multitask alternative without sacrificing ASR performance.},
                         pdf = {https://publications.idiap.ch/attachments/papers/2025/Kumar_IEEEASRU2025_2025.pdf}
}