Federico Ulfo
A community-maintained Hugging Face Space called the Jev Decision Index is tracking, categorizing, and benchmarking every known open-source reproduction of TypeSafe AI's Jev (System One / RLCD) decision model.
The Space, created by @multimodalart, maintains a leaderboard (the Decision Index) that scores each reproduction on a frozen suite of benchmarks, plus a News tab that lists all tracked artifacts color-coded by kind: Decoding (parallel constrained decoding), Diffusion, Trained (Jev-like scoring heads and fine-tunes), Prior art claims, and Explainers. Each card gets a trending score based on engagement (X likes, GitHub stars, Hub likes, views) plus a recency credit so new releases aren't buried.
The Index tab shows every open reproduction that has completed the benchmark suite, with per-category radar charts against Jev itself and a full page per model. The data is generated from a single static JSON bundle; the HTML never needs to change for data updates.
The project also maintains a section on what remains not in the open: TypeSafe's weights, the RLCD algorithm, and any open model matching Jev's calibration claims.
As TypeSafe's Jev sparked a wave of open reproductions, a centralized, objective tracker helps the community compare approaches, identify what works, and measure how close the open ecosystem is getting to the proprietary original. It also documents what pieces of the Jev system are still closed, guiding research priorities.
Sources: Source 1
