Have you ever thought what your abstract would look like as a brain map? Think no longer, NeuroVLM has got you covered!

Methods: Each point is an abstract positioned in 2D from NeuroVLM projected text embeddings, colored by HDBSCAN cluster, and paired with a NeuroVLM-generated brainmap preview. Cluster names are derived from all cluster title sets together using an OpenAI model prompted to return short, globally distinctive theme labels and to disambiguate duplicate names when needed. Hover to scan, click to hold a card open, and use the three filters to narrow the landscape.

Faint links connect nearest semantic neighbors in the UMAP layout. Uncategorized marks abstracts that HDBSCAN did not assign to a stable cluster.

Reference: brainmap previews are generated with NeuroVLM.