@inproceedings{fu-etal-2026-nemotron, title = "Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding", author = "Fu, Yonggan and Whalen, Lexington and Garg, Abhinav and Wu, Chengyue and Khadkevich, Maksim and Oswald, Nicolai and Xie, Enze and Egert, Daniel and Turuvekere Sreenivas, Sharath and Diao, Shizhe and Yu, Chenhan D. and Yu, Ye and Chen, Weijia and Norouzi, Sajad and Liu, Jingyu and Lan, Shiyi and Zhu, Ligeng and Wang, Jin and Jiang, Jindong and Mardani, Morteza and Maghoumi, Mehran and Han, Song and Juki{\'c}, Ante and Tajbakhsh, Nima and Kautz, Jan and Molchanov, Pavlo", editor = "Lee, Hung-Yi and Dev, Sunipa and Li, Jessy and Elliott, Desmond", booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing", month = oct, year = "2026", address = "Budapest, Hungary", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2026.emnlp-main.2416/", pages = "54868--54887", ISBN = "979-8-89176-459-0" }