Awesome Jev

index / MrGKanev/lets-make-jev

lets-make-jev

Jev tests a simple idea: when an LLM only needs to choose a category, score the possible first answer tokens instead of generating a text response. The client maps single-token labels (A, B, …) back to application categories and routes uncertain decisions to human review.

Research and EvalsRoutingReviewPython

Repository

Stars0
LanguagePython
License—
Created2026-10-05
Last pushed2026-11-14
Officialno

Judgment

Genuine Jev project0.35
gate 0.5
Uses Jev at runtime0.77
Is a meta list0.04
Is a reimplementation0.83

Quality scales

Substance1.98 / 3
gate 0.5
Docs2.63 / 3
Novelty2.17 / 3
Composite score0.73

Category distribution

Research and Evals1.00
SDKs and Clients0.00
Integrations0.00
Agent and Dev Tooling0.00
Applications0.00
Games and Simulation0.00
Learning0.00
Other0.00
Assigned categoryResearch and Evals
Category probability1.00
Category confidence1.00
PatternRouting
Pattern confidence0.87

Curation

StatusReview
Reasongenuine 0.35 below 0.5
README pickno
Overridenone

Sources

  • search:topic:jev-ai

Provenance

Modeljev-1.14.0
Question setv2
Judged at2026-11-15 13:24 UTC