The field of language models is moving towards more efficient sampling and decoding methods. Recent research has focused on developing innovative techniques to accelerate sampling from masked diffusion models, improve speculative decoding, and enhance model performance in data-scarcity scenarios. These advancements have the potential to significantly improve the performance and efficiency of language models, enabling them to tackle complex tasks with greater accuracy and speed. Notable papers in this area include Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking, which introduces a novel sampler that accelerates sampling from state-of-the-art models, and Out-of-Vocabulary Sampling Boosts Speculative Decoding, which presents a method for effectively recovering acceptance rates in speculative decoding. Other notable papers include EpiCoDe, Guided Speculative Inference for Efficient Test-Time Alignment of LLMs, and Accelerated Test-Time Scaling with Model-Free Speculative Sampling, which all contribute to the development of more efficient and effective language models.