Jev directory
CommunityJev Models & Reimplementations
Laya Coding Decider v2
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.
- Reading time
- 2 min
- Published by
- Community
- Author
- Evryvilanislemn
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Fine-tuned from convaiinnovations/laya on about 9k coding-decision rows derived from SWE-Gym real GitHub issues and gold fix diffs.
- Answers file keep/drop relevance, priority scores and an injection noul in one predict call over state and questions.
- Held-out results over 1,150 rows: 0.833 overall accuracy, ECE 0.147, Brier 0.086 and 0.411 score MAE.
- Relevance-only accuracy is 0.907 against a roughly 0.64 majority-drop baseline, at a reported p50 latency of 126.4 ms.
- Apache-2.0, loaded with the laya package via laya.load and agent.predict.
Quickstart
python
import laya
agent = laya.load("evryvilanislemn/laya-decider-v2")
result = agent.predict(state, questions)
print(result["answers"])Watch out
Card reviewed 2026-10-17; the 0.833 accuracy, 0.907 relevance accuracy and 126.4 ms p50 are self-reported by the author on a 1,150-row held-out split, the card does not document the hardware behind the latency, and no third-party reproduction is linked.
Questions about this entry
Short answers drawn from the linked project or page.
A coding decision model fine-tuned from convaiinnovations/laya on about 9k coding-decision rows derived from SWE-Gym, covering keep/drop file relevance, priority scores and an injection noul question.
About 9,000 coding-decision rows derived from SWE-Gym/SWE-Gym, built from real GitHub issues and their gold fix diffs, so the decisions track realistic file-relevance and prioritisation patterns.
On 1,150 held-out rows: 0.833 overall accuracy, ECE 0.147, Brier 0.086, score MAE 0.411 and relevance-only accuracy 0.907 against a majority-drop baseline of about 0.64. All figures are self-reported.
A trivial strategy that always drops would score about 0.64 on the relevance subset; the model's 0.907 is well above that, which is the card's argument that the relevance signal is real.
The card reports a p50 latency of 126.4 ms per prediction, without documenting the hardware or batch size, so treat the figure as indicative rather than a benchmark.
Install the laya package, call laya.load with the checkpoint name, then agent.predict(state, questions) and read result["answers"]. The card shows exactly this four-line usage block.
The evaluation is a single held-out split of 1,150 rows reported by the author; the card publishes no hardware for the latency, no confidence intervals and no independent replication, and the model stays within coding decisions.
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 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 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
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 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.
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.