openJev-verdict 2.0: a 151M calibrated decision engine

A non-autoregressive decision engine on ModernBERT (151M) with calibrated uncertainty, an audit of the Jev benchmark, and an in-browser WebGPU playground; Apache-2.0 with PyTorch and ONNX weights on Hugging Face.

Notes

Author-run on LocalLLaMA/typed-decisions: 77.10% accuracy, 0.0636 Brier, 0.0144 ECE, described as beating Jev and Laya on that set; a reshare calls it '>7x faster than Jev'.

source-backed — Public repository, docs or live artifact. About this label

Destinations

Also linked

• Verdict-open-jev (v1) ↗

Filed under

• #replica
• #modernbert
• #calibration
• #webgpu
• #open-source

In collections

• Local & replica models →