modality-adaptive-reasoning-visualizations

Enable small vision-language models to reason over diverse data types by converting latent embeddings into visual representations, achieving specialized performance without domain-specific training.

Independent GitHub snapshot · Automated classification · Original documentation governs