Jev directory
CommunityJev Models & Reimplementations
Qwen3-ASR-0.6B-Jev-Dutch
Apache-2.0 fine-tune of Qwen/Qwen3-ASR-0.6B published on Hugging Face that adds a Jev head for Dutch speech, tagged for data curation, speaker attributes, and audio quality. Ships custom modeling code and 782M BF16 parameters in one safetensors file.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- Fine-tune of Qwen/Qwen3-ASR-0.6B carrying the jev and qwen3_asr_jev_dutch tags on Hugging Face.
- 782,502,987 BF16 parameters in a single unsharded safetensors file of roughly 1.57 GB.
- Trained on Common Voice 17.0, VoxPopuli, Multilingual LibriSpeech, and speechocean762.
- Loads through custom code: transformers maps it to modeling_jev.JevModel via AutoModel.
- Tagged with arXiv 2610.21614, Qwen-Audio-Jev: A Unified Interface for Audio Decisions.
Questions about this entry
Short answers drawn from the linked project or page.
It is a Hugging Face model repo, tvosch/Qwen3-ASR-0.6B-Jev-Dutch, built as a fine-tune of Qwen/Qwen3-ASR-0.6B with a Jev head. Its pipeline tag is audio-classification and its other tags cover speech, data curation, speaker attributes, and audio quality for Dutch audio.
The repo declares custom code, with transformersInfo listing AutoModel and the custom class modeling_jev.JevModel. Loading therefore requires the transformers library with remote code enabled, and the repo also documents a pipeline helper for high-level use.
The base model is Qwen/Qwen3-ASR-0.6B, recorded as both base_model and base_model:finetune. The safetensors metadata reports 782,502,987 BF16 parameters, unsharded, with a total file size of about 1.57 GB and no quantization.
The only language tag is Dutch (nl). Four datasets are listed: mozilla-foundation/common_voice_17_0, facebook/voxpopuli, facebook/multilingual_librispeech, and mispeech/speechocean762.
The repository is tagged license:apache-2.0. That is the license recorded on the model page; anyone planning redistribution or commercial deployment should read the model card and the base model's own terms before relying on it.
The repo is not gated and no inference providers are enabled, so there is no hosted endpoint to call. The weights are public safetensors, roughly 1.57 GB in BF16, which is small enough to download and run on local hardware.
It is scoped to Dutch speech and to audio-classification style tagging such as data curation, speaker attributes, and audio quality. For other languages, for general text generation, or for tasks outside audio classification, another model would be the better starting point.
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
AgriJev: Efficient Agricultural Candidate Scoring with Multi-Scale Adapted Models (arXiv 2610.19478)
A family of adapted Qwen3.5 models from 0.8B to 9B that scores agricultural candidates in one forward call: AgriJev-9B reaches 75.44% mean accuracy across 17 test views and narrows the gap to hosted Jev 1.13.0.
Jev Models & ReimplementationsArticle#community#research#arxiv
Community
Open Jev-style decision model built on LiquidAI's LFM2.5-350M backbone, a clone of kev. Serves the System One API contract so the TypeSafe SDK works against a local server, returning probabilities for typed questions from one forward pass.
Jev Models & Reimplementations#open-models#python#local
Communityfranckverrot/lev
Open-weight vision-language judge that takes a task instruction plus 10 frames of a robot manipulation attempt and predicts success or failure with calibrated probabilities in a single forward pass. Released with inference code and a 20,628-prediction evaluation set.
Jev Models & ReimplementationsArticle#arxiv#open-models#evaluation
Community
arXiv paper introducing Iris, a System One design that accepts image, text, or mixed candidates while leaving Jev's decision module untouched, with order-invariance, null-probability, and candidate-reuse results on a Multimodal-Mind2Web benchmark.
Jev Models & ReimplementationsArticle#arxiv#open-models#calibration
Community
arXiv paper testing whether the readout and adaptation choices used for direct selection still hold up in cross-request entailment reranking for JEV decision models. It combines a mean-span readout with selective backbone adaptation and reports accuracy on MC11, an average over eleven reranking configurations.
Jev Models & ReimplementationsArticle#arxiv#research#classification
Community
Hugging Face repository hosting GGUF weights for Jev-Style-2B-Decision-v4, a 2B Jev-style decision model published under the chaoliangUNSW account. The listing includes a chat template that renders messages, reasoning content, tool calls, and vision inputs.
Jev Models & Reimplementations#open-models#local#calibration
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.