ELF in 1 Minute: Language Diffusion in Continuous Embeddings

Most diffusion language models work over discrete tokens. ELF takes a different route: keep language in continuous embedding space, run Flow Matching there, and decode back to tokens only at the final step. Paper: ELF: Embedded Language Flows arXiv: https://arxiv.org/abs/2605.10938 #AI #MachineLearning #DiffusionModels #LanguageModels #FlowMatching #ResearchPaper Most diffusion language models work over discrete tokens. ELF takes a different route: keep language in continuous embedding space, run Flow Matching there, and decode back to tokens only at the final step. Paper: ELF: Embedded Language Flows arXiv: https://arxiv.org/abs/2605.10938 #AI #MachineLearning #DiffusionModels #LanguageModels #FlowMatching #ResearchPaper