index / clduab11/jev-test
jev-test
Pre-registered benchmark: can a 2B local model (Gemma 4 E2B) answer web questions without making things up when a decision model (TypeSafe Jev) makes every call? SearXNG for search, MemPalace for verbatim memory, seven arms including open local judges. Spec and thresholds fixed before any run.
Research and EvalsVerificationListedPython
Repository
Stars0
LanguagePython
LicenseMIT
Created2026-09-19
Last pushed2026-09-19
Officialno
Judgment
Genuine Jev project0.97
gate 0.5
Uses Jev at runtime0.93
Is a meta list0.03
Is a reimplementation0.07
Quality scales
Substance2.02 / 3
gate 0.5
Docs2.83 / 3
Novelty2.55 / 3
Composite score0.79
Category distribution
Research and Evals1.00
SDKs and Clients0.00
Learning0.00
Other0.00
Integrations0.00
Agent and Dev Tooling0.00
Games and Simulation0.00
Applications0.00
Assigned categoryResearch and Evals
Category probability1.00
Category confidence1.00
PatternVerification
Pattern confidence0.82
Curation
StatusListed
Reasongate passed
README pickno
Overridenone
Sources
- search:jev typesafe
- search:topic:jev
- search:topic:system-one
- search:jev in:name,description,readme created:2026-09-19
- code:"api.typesafe.ai"
- code:"jev-latest"
Provenance
Modeljev-1.13.0
Question setv2
Judged at2026-09-20 11:23 UTC