CommunityJev Models & Reimplementations
DepGuard Jev-Style 1.5B
A Qwen2.5-1.5B-Instruct fine-tune that emits typed dependency verdicts, safe, caution, or block with a recommendation, for Dependabot and Renovate pipelines, reporting 95.7% F1 on the card's 4,800-update OSV-Bench.
- Published by
- Community
- Author
- JackKozmo29
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- 1.5B-parameter fine-tune of Qwen2.5-1.5B-Instruct returning one JSON verdict per package update.
- OSV-Bench: 95.7% F1 (96.0 precision, 95.4 recall) on 4,800 CVE-annotated package updates.
- Malicious-package holdout: 95.1% block rate with 4.9% false positives, above hosted Jev's 94.3%.
- LoRA r=16, alpha=32 on 92,000 PyPI, npm, Maven, and Go package updates from 2018-2024.
- Card table: ~82ms local decisions on MPS versus ~125ms for the hosted Jev API, free after download.
Watch out
Apache-2.0, with the Qwen2.5-1.5B-Instruct base under its own Qwen licence. All benchmarks and latency figures are the card's own runs, and its quickstart code loads enterprise-ai-lab/depguard-jev-style-1.5b rather than the JackKozmo29 repo path. Verdicts use package metadata and descriptions, not static analysis of contents, so clean-metadata malicious packages can score safe.
Questions about this entry
Short answers drawn from the linked project or page.
One JSON object per package update with verdict (safe, caution, or block), a one-sentence reason, cve_count, and recommendation (upgrade, pin, replace, accept, or reject). The card shows a lodash 4.17.15 to 4.17.21 review returning safe with cve_count 1.
LoRA with rank 16 and alpha 32 over 12 epochs on 92,000 package updates from PyPI, npm, Maven, and Go modules from 2018-2024, using the OSV database, GitHub Advisory Database, and manual security-team labels as ground truth.
95.7% F1, with 96.0% precision and 95.4% recall on 4,800 CVE-annotated updates. The card places that just under the hosted Jev API's 96.1% F1 and above GPT-4o-mini's 89.9% and OSV-Scanner's 82.7%.
On supply-chain attack samples it blocks 95.1% with a 4.9% false positive rate, while OSV-Scanner blocks 61.2% with 1.1% false positives. The card notes rule-based tools miss novel malicious packages not yet in the CVE database.
The verdict is based on metadata and description, not static analysis of package contents, and novel malicious packages with clean metadata may score safe. The card says to combine it with OSV-Scanner for defense in depth.
Yes per the card: serve it as an OpenAI-compatible endpoint and point ChatOpenAI at localhost:8000/v1 with model depguard-jev-style-1.5b. It also allows LoRA fine-tuning on an internal allowlist, unlike the hosted API.
The Hugging Face repo is JackKozmo29/depguard-jev-style-1.5b, and the card presents it as a drop-in local alternative to TypeSafe Jev rather than an official TypeSafe model. Its quickstart instead loads enterprise-ai-lab/depguard-jev-style-1.5b, so verify which weights you pull.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
YouTube: Kubernetes admission control with a decision model, fail-closed.
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
Superagent's security triage model: a continual fine-tune of AutoJev-27B that returns calibrated probabilities for prompt-injection and content checks, with a release policy at 0.70 unsafe.
Jev Models & Reimplementations#community#open-models#local
Community
akhilaaa3's merged multimodal decision classifier on Gemma 4 12B IT: probabilities for text, image, audio or video options, 87.57% on DecisionBench Medium and MMAU 63.10%, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Mithril's open, Jev-shaped typed-decision model on DeBERTa-v3-large: choice up to 255 options, score and noul from one forward pass, 0.854 in-domain with ECE 0.022, trained on public gold labels only.
Jev Models & Reimplementations#community#open-models#local
Community
TokenRhythm's 4B prefill-only decision model built on NeoHorse-1-4B: given a state and application-defined questions it returns Choice, Noul, and Score probabilities without generating text, reporting a 77.70 six-group aggregate under Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
ggmlc-compiled GGUFs of the Laya typed-decisions specialist on ModernBERT-large: F16, Q8_0, and UD_Q4_K_M files for invoice, security, customer-service, and harness presets, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
An EmbeddingGemma 2 fine-tune that answers choice, score, and noul questions from a state and a plain-data schema, with a 0.5M-parameter schema-conditioned scorer over 768-d embeddings.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.
Jev Models & Reimplementations#open-models#training#local
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
Read article
From the Jev community
Posts from builders shipping with Jev right now.