Explore text diffusion over continuous or learned representations. Check whether “latent” means word embeddings or an autoencoder-compressed representation, and how the model maps it back to text.
How to read and compare
What representation is denoised, and how is it decoded or rounded?
Are representation and decoding errors separated from generation quality?
Diffusion-LM studies continuous word vectors; it does not validate every compressed-latent text model.
A code link does not guarantee reproducibility. Check the original dataset, metric, protocol and conditions.
Language remains an outlier in generative modeling: while images, video, and audio are increasingly modeled in continuous latent spaces, text generation still relies predominantly on discrete tokens. Existing continuous language models either inherit embedding spaces not designed for joint generation and decoding, or compress autoencoded latents to ease diffusion, sacrificing token-level fidelity. Instead of simplifying the representation to suit the generative model, we preserve a high-capacity, decodable text latent and design the diffusion model to learn its distribution directly. We intr
Useful for: Not assessed · Limitation: Not assessed
Handwritten text generation (HTG) conditioned on writer style has been widely studied for Latin scripts, but remains underexplored for low-resource and non-Latin writing systems, leaving open how well existing models generalise beyond the Latin domain. Cyrillic, particularly Ukrainian, lacks both large-scale writer-labeled datasets and empirical evidence of such generalisation. To address this gap, we construct a Ukrainian handwritten word dataset of 126,177 images from 308 writers using connected-component segmentation, quality filtering, and targeted oversampling of underrepresented Ukrainia
Useful for: Not assessed · Limitation: Not assessed