{"slug": "flow-duality-and-source-geometry-for-categorical-generation", "title": "Flow Duality and Source Geometry for Categorical Generation", "summary": "A new arXiv paper (2609.10863v1) identifies a duality between continuous and discrete flow matching, showing that projecting continuous convex-interpolant paths with one-hot targets through a position-wise argmax yields discrete convex-interpolant paths. The result requires source laws with coordinate symmetry and boundary regularity, making the continuous source distribution an explicit design choice for categorical generation. The authors derive induced discrete interpolation behavior for Gaussian, bounded-uniform, and centered negative-exponential sources, finding that different source geometries produce qualitatively different transition timing and vocabulary-size dependence, with small visual diagnostics and a short language-modeling pilot suggesting these source-design effects appear in learned transports and early generative quality.", "body_md": "arXiv:2609.10863v1 Announce Type: new \nAbstract: Continuous and discrete flow matching are usually treated as separate constructions. This paper identifies a duality between them: projecting continuous convex-interpolant paths with one-hot targets through a position-wise argmax yields discrete convex-interpolant paths. The result requires source laws with appropriate coordinate symmetry and boundary regularity, and it makes the continuous source distribution an explicit design choice for categorical generation. We derive the induced discrete interpolation behavior for Gaussian, bounded-uniform, and centered negative-exponential sources, showing that different source geometries lead to qualitatively different transition timing and vocabulary-size dependence. Small visual diagnostics and a short language-modeling pilot suggest that these source-design effects can also appear in learned transports and early generative quality.", "url": "https://wpnews.pro/news/flow-duality-and-source-geometry-for-categorical-generation", "canonical_source": "https://arxiv.org/abs/2609.10863", "published_at": "2026-09-11 04:00:00+00:00", "updated_at": "2026-09-11 04:28:47.008492+00:00", "lang": "en", "topics": ["machine-learning", "generative-ai", "ai-research", "large-language-models"], "entities": ["arXiv", "Flow Duality", "Source Geometry", "Categorical Generation"], "alternates": {"html": "https://wpnews.pro/news/flow-duality-and-source-geometry-for-categorical-generation", "markdown": "https://wpnews.pro/news/flow-duality-and-source-geometry-for-categorical-generation.md", "text": "https://wpnews.pro/news/flow-duality-and-source-geometry-for-categorical-generation.txt", "jsonld": "https://wpnews.pro/news/flow-duality-and-source-geometry-for-categorical-generation.jsonld"}}