The method of finding a root in mathematics, based on repeated division of a segment in half and the subsequent selection of a subinterval in which the root is thought to be located.
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).