Jev directory
CommunityJev Models & Reimplementations
Laya Typed-Decisions Fine-Tune
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.
- Reading time
- 2 min
- Published by
- Community
- Author
- huangxn
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Trained on the 1,200 training cases (6,000 decisions) of the independent LocalLLaMA/typed-decisions benchmark.
- Reports 0.777 accuracy on the official 400-case test set, above TypeSafe Jev 1.13.0 at 0.727 and the teacher self-agreement ceiling at 0.735.
- Head-to-head: Brier 0.050 versus Jev's 0.148, score MAE 0.212 versus 0.391, and 0.995 within one level versus 0.952.
- 151.4 ms p50 self-hosted at $0.00 per case versus 710 ms and $0.0004 for the hosted Jev comparison.
- Covers Agent Trace Observability, Customer Service, Invoice Processing and Security Incidents workflows; Apache-2.0.
Quickstart
python
import laya
agent = laya.load("huangxn/laya-typed-decisions")
result = agent.predict(state, questions)
print(result["answers"])Watch out
Card reviewed 2026-10-17; every figure comes from the fine-tune's own head-to-head table, where the Jev 1.13.0 comparison was run by the same publisher rather than by TypeSafe, and the model trains and is evaluated on splits of the same benchmark.
Questions about this entry
Short answers drawn from the linked project or page.
A Laya checkpoint fine-tuned on the 1,200 training cases (6,000 decisions) of the independent LocalLLaMA/typed-decisions benchmark, answering workflow states and typed questions in a single forward pass.
On the official 400-case test set the card reports 0.777 accuracy versus 0.727 for TypeSafe Jev 1.13.0, with Brier 0.050 versus 0.148, score MAE 0.212 versus 0.391 and 0.995 within one level versus 0.952.
The benchmark defines a ceiling of 0.735, the rate at which its teacher labels agree with themselves; the fine-tune reports surpassing that ceiling with 0.777 accuracy on the 400-case test set.
Four: Agent Trace Observability, Customer Service, Invoice Processing and Security Incidents, for 2,000 decisions in total across the official 400-case test set.
Install the laya package, call laya.load with huangxn/laya-typed-decisions, then agent.predict(state, questions) and read result["answers"]; the card shows this minimal four-line usage block.
Apache 2.0, per the card, developed by Convai Innovations. The training data is the typed-decisions benchmark, so check that benchmark's own terms if you redistribute data or derivatives.
The comparison numbers are the publisher's own runs rather than an independent reproduction, the model trains and is tested on splits of the same benchmark, and the card presents no confidence intervals for the 400-case results.
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
AutoTrust's combined System 1 and System 2 bundle on Gemma-4-26B-A4B-it: Decision Index 0.2.1 of 58.05, a 24-slot head with noul, choice and score, and the base chat model left bit-identical.
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 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
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.