verse-embedding-visualization-documents

Optimize vision-language models for document tasks via embedding visualization and clustering-guided data generation. Identify error-prone regions in visual space and synthetically augment training data targeting weak areas.

Independent GitHub snapshot · Automated classification · Original documentation governs