jev-acento
Language evaluation: pre-registered paired audit of Jev on Spanish over 3,200 human-labelled items, finding that a Spanish `state` costs 3.0-6.4 pp of accuracy and roughly doubles ECE on XNLI and PAWS-X while writing `instructions` in Spanish changes nothing, and shipping a CLI to rerun the same comparison on your own labelled data.

- GitHub stars
- 0 as of 2026-09-22
- Forks
- 1
- Language
- Python
- License
- MIT
- Created
- 2026-09-20
- Last updated on GitHub
- 2026-09-21

What Jev does in jev-acento
The project describes itself this way: Language evaluation: pre-registered paired audit of Jev on Spanish over 3,200 human-labelled items, finding that a Spanish `state` costs 3.0-6.4 pp of accuracy and roughly doubles ECE on XNLI and PAWS-X while writing `instructions` in Spanish changes nothing, and shipping a CLI to rerun the same comparison on your own labelled data. In the Community additions category, Jev typically decides a single well-defined judgment, such as a label, a score or a yes/no check.
Verification
Verified. We confirmed this repository and its Jev relevance through the GitHub API and the source lists below. A specific file link has not been recorded yet.
Topics
- benchmark
- calibration
- expected-calibration-error
- jev
- llm-evaluation
- nlp
- pre-registration
- reproducible-research
- spanish-nlp
- typesafe-ai
How Community additions projects use Jev
Projects added from community lists and direct submissions. Many are small, single-purpose experiments, which makes them some of the easiest places to see how a Jev question is framed in real code.
Related Jev projects in Community additions
- Life Chess × JevARCJ137442/jev-lifeExperimental game design: a turn-based Conway board where each side's move is one Jev `Boolean` per legal cell in a single request, with no heuristic fallback and a user-set confidence threshold flagging unsure turns; the game is new, so there is no established play to copy, and its rules are not self-evident — they render from editable templates with auto-filled placeholders, so a designer can rewrite a rule and re-ask.★ 0
TypeScript - PlotVeilDearest/plotveilSpoiler protection: Chrome extension that covers each YouTube comment while one Jev `Noul` question, batched 20 at a time, answers whether it reveals a concrete plot event of the video being watched or of another title the user protects, with the extension owning the 0.85 / 0.7 / 0.5 threshold and keeping the comment covered when the check fails.★ 0
TypeScript
jev-table-import-mapperDuvInc/jev-table-import-mapperData import: maps an uploaded CSV's columns onto a destination table with a strict deterministic name-equality pass, then one Jev `Noul` per remaining (source, destination) pair plus a guard `Noul` per incoming column, mapping 10 of 10 columns of a 23-column export at 253 questions in one call, 915 ms, $0.0012, unmapped fields left visible above a 0.75 threshold rather than guessed.★ 0
JavaScript- Diffusion JevHangzhi/diffusion-jev-sglangVisual classification: independent Jev-style DiffusionGemma/SGLang server that selects doodle and flower labels from image pixels with typed Choice questions and displays candidate scores in a drawing playground, with public evaluation artifacts and uncalibrated probabilities.★ 0
Python - jev-evalShogo-nfrealmusic/jev-evalThird-party check: compares Jev against GPT-4o-mini and Claude Sonnet 4.5 under identical conditions on the same judgment task.★ 0
TypeScript
Feed LensSkywalkerDarren/feed-lensSocial media: uses Jev `Noul` judgments against per-platform, user-defined topic and expression labels to annotate Weibo, Threads and X posts directly in a Chrome extension.★ 0
JavaScript