@article{cactusacceleratingautoregressivedecoding, title = {Cactus: Accelerating Auto-Regressive Decoding with Constrained Acceptance Speculative Sampling}, author = {Yongchang Hao and Lili Mou}, year = {2026}, eprint = {2604.04987}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2604.04987}, }