jevVIDEO
Explaining the Jev-style model: scoring a menu rather than writing an answer
A video explains Jev's form: context plus N options produce one probability each in a single forward pass with no autoregression, which is why it can be 20-200x faster.
Md Ismail Šojal 🕷️
@0x0SojalSecArchived 2026.09.18Originally posted 2026.09.17
Md Ismail Šojal 🕷️ @0x0SojalSecOriginally posted 2026.09.17
Train your own jev like model. - Jev doesn’t write the answer. It scores the menu. - Context + N options to one probability each. One pass, No autoregression. That’s why 20–200× faster exists as a product shape. A classifier that takes its categories as input.
This post quotes another post: x.com/0x0sojalsec/status/2100386882979066015
x.com/0x0sojalsec/status/2100388850917118267Archived 2026.09.18 10:16