Research on Jev: a map of the first preprints

What has been verified in this overview

This is a map of the preprints we have identified. We read the abstracts; we have not independently replicated the full methods and results. Including a paper here is not a recommendation for production use.

Jev for Scientific Decisions

The paper of 21 September 2026 examines the selection of semantic relationships before a deterministic calculation. The abstract reports twenty Choice questions across ten scientific cases and twelve model configurations. It is a small, purpose-built test, not a broad validation of a scientific assistant. S22

Visual Jev

The paper of 22 September 2026 examines sharing visual context across multiple questions about an image. It is a separate research proposal; the title does not announce image input for TypeSafe’s hosted Jev. S23

Jev in edge orchestration

Two papers by Delong Li and co-authors from 19 September 2026 examine the interpretation of requests and the subsequent planning of computations. They combine model decisions with numerical planning; one explicitly describes live measurements and simulation, the other also a real two-node service. S24 S25

Our editorial interpretation

These topics point to interesting directions of use, not to a settled consensus on quality. We do not automatically treat papers by the same authors as two independent confirmations. The next stage requires reading the methodology, the available code, the data and the negative results.

When JEV.sk creates its own benchmark, it will have a separate dataset, code and protocol. Until then, no fictitious experiments or charts with made-up measurements will be added to the site.