SciNCL is a specialist — it lives and breathes scientific literature, having been trained on citation networks from Semantic Scholar. It produces dense embeddings that capture the semantic relationships between research papers with notable precision, understanding domain-specific language that general-purpose models often flatten or miss. Its narrow focus is both its strength and its boundary: outside academic text, it has little to offer.