Jeva is a decision model architecture that encodes options independently to support order stability, variable candidate sizes, precomputed caching, and distinct detection of absent answers.
Qi Li and Xinchao Wang’s architecture stores options separately, then updates the question against those memories. This makes option order irrelevant by construction and allows reuse across questions. A separate head estimates whether the answer is absent; it is not a general uncertainty detector. The reported 27.98× median speedup assumes full option reuse over 50 questions. No released code or weights were established.
How you can use it
Borrow the paper’s design when building a service that repeatedly chooses from the same large menu. It stores each option separately so new questions can reuse that work. This is an architecture to implement, not a verified downloadable app; its separate no-match check is different from merely having low confidence.