Before you press play
What you’ll find in the video
- Jev evaluates content using three core primitives: Noul for boolean probabilities (0 to 1), score for rubric-based grading, and choice for selecting from discrete categories.
- Multiple questions using different primitives can be batched together to evaluate a single state input in one request.
- Probability and confidence metrics enable automated conditional logic, routing decisive cases automatically while flagging borderline results for human review.
Worth knowing
Auto-generated English captions reviewed with Gemini. Fast typed outputs and reported sub-500ms latencies are vendor demonstrations, not formal controlled benchmarks, and typed outputs do not guarantee correct decision logic.