Jev directory
CommunityJev Models & Reimplementations
Turnike-JEV-4B
A bilingual Turkish and English 4B Jev-class decision model from Berkay Hasret: a roughly 130 MB LoRA adapter on Qwen/Qwen3.5-4B that answers up to 16 typed questions per text in one pass and reports 0.774 JevBench macro accuracy.
- Reading time
- 3 min
- Published by
- Community
- Author
- berkay123001
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- LoRA adapter (r=16, about 130 MB) on Qwen/Qwen3.5-4B; up to 16 questions about one text in a single forward pass.
- Reports 0.774 macro accuracy on full JevBench (22 configs, 22,773 rows) with macro ECE 0.084 from raw probabilities.
- tr_insan Turkish: 0.748 vs 0.727 for Jev, 0.714 for Microsoft-Decision-1 (p = 0.003), 0.708 for Kev-4B and 0.675 for Intern-Decision-4B.
- About 109 ms per decision in bf16 and 82 ms in FP8; the FP8 copy is 5.5 GB on disk and about 4.5 GB of GPU memory.
- Shares one recipe with Turnike-JEV-0.8B and 9B (macro 0.681 / 0.774 / 0.783); two training-data shortcuts were found and fixed.
Quickstart
python
# pip install torch "transformers>=5.17" peft accelerate ; karar.py: hf download berkay123001/turnike-jev-4b karar.py --local-dir .
from karar import Karar
k = Karar("berkay123001/turnike-jev-4b")
# EN
k.sor("The production database has been down since 9am. Please escalate now.",
{"Does this message require immediate action?": {"type": "noul"},
"Which team should handle it?": {"billing": "payments, refunds", "infra": "servers, outages", "sales": "pricing"}})Watch out
Card reviewed 2026-10-17; the 0.774 JevBench, 0.748 Turkish, 0.084 ECE, latency and cost figures are self-reported by the author and the run was not submitted to the public leaderboard, so the small 0.004 lead over the top entry is not independently verified. The training datasets are not released, and gold labels for much of the Turkish evaluation come from labelling rather than human sources on some splits.
Questions about this entry
Short answers drawn from the linked project or page.
An open, bilingual Turkish and English 4B Jev-class decision model by Berkay Hasret: a LoRA adapter (r=16, about 130 MB) on Qwen/Qwen3.5-4B. Text plus closed questions go in, calibrated option probabilities come out, fully offline.
Macro accuracy 0.774 on the full JevBench test (22 configs, 22,773 rows), above the best 4B leaderboard entry the card cites (0.751) but below the top entry overall (0.770, a 35B MoE). The card notes the 0.004 lead is small and the run was not submitted.
On 1,379 human-labelled Turkish decisions it reports 0.748, against 0.727 for Jev 1.13.0, 0.714 for Microsoft-Decision-1 (paired McNemar p = 0.003), 0.708 for Kev-4B and 0.675 for Intern-Decision-4B.
About 109 ms per decision in bf16 and 82 ms in FP8, with eight questions on one text in one pass at about 156 / 126 ms. The adapter is about 130 MB; the FP8 copy is 5.5 GB on disk with about 4.5 GB GPU memory. No API is required, and CPU works but slowly.
Macro ECE is 0.084 from raw probabilities on full JevBench, against 0.113 for Jev 1.13.0, and the card notes leaderboard models that fit a recipe reach about 0.05, so you can fit a temperature if you need tighter calibration.
Berkay Hasret built it as a LoRA on Qwen/Qwen3.5-4B with the same data, draw and hyperparameters as the 0.8B and 9B siblings. It was not distilled from a commercial decision API, and the card declares AI assistance with the card and code.
Two training-data shortcuts were found and fixed, the training data is not released, and the card lists which test sets were used to pick the run. The 'none of the above' behaviour is checked on a balanced set whose option count does not give the answer away.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
A 9B bilingual Turkish and English Jev-class decision model from Berkay Hasret: a 173 MB LoRA adapter on Qwen/Qwen3.5-9B that answers up to 16 typed questions about one text in a single forward pass and reports 0.783 macro accuracy on full JevBench.
Jev Models & Reimplementations#community#open-models#local
Community
A LoRA plus 24-slot decision head over Qwen3.8-27B from Ming Xu's train-jev project: it serves TypeSafe's System One API with 0.851 Kev transfer-v4 accuracy and ECE 0.053, and needs an 80 GB GPU.
Jev Models & Reimplementations#community#open-models#local
Community
A 24 GB-GPU alternative from the train-jev project: Qwen3.5-4B plus a rank-16 LoRA and 24-slot decision head scoring 0.647 PK accuracy, 40.4 Decision Index and 43 ms p50 latency, ahead of Kev-4B at the same size.
Jev Models & Reimplementations#community#open-models#local
Community
A Laya fine-tune for coding decisions: roughly 9k rows from SWE-Gym teach keep/drop file relevance, priority scoring and injection detection, reporting 0.833 accuracy on 1,150 held-out rows at 126 ms p50.
Jev Models & Reimplementations#community#open-models#local
Community
A Laya fine-tune from Convai Innovations on the 1,200-case typed-decisions benchmark: it reports 0.777 accuracy on the official 400-case test set, above TypeSafe Jev 1.13.0 at 0.727 and the teacher ceiling at 0.735.
Jev Models & Reimplementations#community#open-models#local
Community
Mapika's NVFP4 fine-tune of Gemma-4-31B-it answers typed Choice, Noul and Score questions with calibrated probabilities and no generated text, self-reporting a 64.27 chance-corrected public Decision Index and 39 ms median latency on a B300.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.