第 11 份独立评测：克罗地亚语零样本 vs 自己微调的 BERT（Senko Rašić，09-22）
Independent evaluation by Senko Rasic comparing Jev zero-shot on Croatian text against a self-fine-tuned BERT.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
A blog post by Senko Rasic, listed as the 11th independent evaluation of Jev, that compares Jev used zero-shot on Croatian-language classification against a BERT model the author fine-tuned himself.
It is useful for readers asking whether a hosted decision model can replace a small task-specific classifier in a lower-resource language. The decision model is what is being measured.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | How does zero-shot Jev compare with a fine-tuned BERT on Croatian classification?(inferred) | — |
How it works
- Pick a Croatian classification task
- Run Jev zero-shot with typed questions
- Fine-tune a BERT baseline on the same task
- Compare results
Strengths
- Third-party comparison against a practical baseline
- Covers a non-English language
Caveats
- Only a short list summary was available; results not reproduced here
- Single-author evaluation
At a glance
- Type
- Article
- Status
- research
Use cases and integrations
How curated lists describe it
The 11th independent evaluation: zero-shot Croatian vs.
wh000wh000/awesome-jev-live
Frequently asked questions
What is compared?
Jev used zero-shot versus a BERT model fine-tuned by the author, on Croatian text.
Who wrote it?
Senko Rasic, on blog.senko.net.
Where are the results?
In the article at https://blog.senko.net/analyzing-jev-a-new-ai-model.
Related projects
Bilingual research report on what Jev is, how its claims compare with independent evidence, and the open-source alternatives.
ProjectresearchJapanese research deck, article and source ledger tracking TypeSafe's Jev and System One models over time.
ProjectdiscussionMarkdownLiving index of real Jev use cases, repos, patterns and benchmarks, with every headline figure tagged self-reported or independent.
ArticleproductionEvidence-aware map of 137 research items on Jev and typed probabilistic decisions, with evidence tiers and an interactive atlas.
ArticleproductionJev project list where each entry links to the line of code calling Jev, plus an independent 2,390-question test of vendor claims.
ArticleproductionDatawhale's Chinese Jev course: notebooks on the three primitives, translated recipes, evaluation and local Laya fine-tuning.
ArticleproductionPython