Wednesday, 14 October 2026 · 6 min read
Jeva architecture enables stable, scalable, and cacheable system one decisions
Jeva introduces a decision-native model that independently encodes requests and options, refines request state conditioned on options, and explicitly models null outcomes. It achieves strong accuracy, perfect permutation invariance, scalable candidate handling, and large latency reductions via caching.
Based on the paper Jeva: System One Models from First Principles by Qi Li, Xinchao Wang
Challenges in system one decision modeling and structural requirements
System one models aim to produce direct, low-latency decisions such as selecting a tool or assigning a label, without generating intermediate explanations. Such decisions involve choosing from a set of candidates, which imposes structural requirements on the model.
These requirements include null-awareness (detecting when no candidate applies), permutation invariance (decisions should not change if candidate order changes), option scalability (handling variable candidate set sizes without changing output space), and cacheability (reusing computations for recurring candidates across requests).
Existing approaches often encode requests and candidates jointly as serialized text, which couples candidate representation with request context, limits reuse, and treats unordered candidate sets as ordered sequences, causing instability and inefficiency.
Sign in to read the whole paper
You have read the opening. The rest is free with a Bezen account, and no card is needed.
In the rest of this paper
- How the method works, step by step, with a flow chart
- Three key formulas from the paper, explained symbol by symbol
- Three figures from the paper, and how to read them
- How it was tested, what was found, and where it falls short