Specter2 Base is a focused, methodical embedder built specifically for scientific literature. It understands the structure of academic papers — titles, abstracts, citations — and produces embeddings that reflect domain relationships rather than surface-level word similarity. It handles the dense, jargon-heavy language of research text without losing semantic coherence, though it's narrowly specialized and won't generalize well outside academic contexts.