Knowledge Constellation

The Knowledge Constellation was built entirely offline from this conference's own submission data, and only presentations whose authors explicitly opted in are included. Each opted-in abstract was turned into a semantic embedding so that papers about similar ideas sit near one another; those embeddings were projected to this 2-D map and grouped into topical clusters, while similarity between papers defines the "related work" links. A local AI model then read each cluster and wrote its theme name and a short "why these connect" note. The finished map, clusters, links, and labels were precomputed into a static file and drawn in your browser — there is no server, database, or live AI. Map distance is approximate, and related work reflects topical similarity, not research quality or a probability of collaboration.