Abstention Experiment

pytest-jev 是一个 pytest 插件，用 TypeSafe 的 Jev 模型为 LLM 应用输出提供语义断言。它不检查确切的词语，而是检查文本的语义含义，返回校准概率并在测试失败时显示违规主张的置信度。

Evaluates decision probabilities on labeled examples and compares accuracy, calibration and automation coverage. It helps choose when to accept Jev decisions or escalate them, and supports regression checks when model behavior changes.

Adds diagnostic and routing judgments to OpenTelemetry log records. Jev helps identify records worth further model analysis while retaining the original archive, separating triage decisions from long-form diagnosis.