Jev directory
CommunityJev Models & Reimplementations
Jev Boost re8 HoVer Train Embeddings (Qwen3.6-35B-A3B-FP8)
Hugging Face dataset that renders HoVer claim-verification items as Jev Decision Index retrieval and classification requests, with Qwen3.6-35B-A3B-FP8 state, action, and hard-negative embeddings for post-training CLM heads.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- 12,021 train pairs and 1,290 validation pairs, each with up to 8 hard negatives.
- Vectors are float16 [n, 2048], produced by vLLM pooling with the last token and L2-normalized.
- States render the benchmark's request plus its instructions and never list the options.
- One-token actions and negatives are zeroed and resolved from a separate single-token embedding dataset.
- Total file size is 170 MB; source text is HoVer CC BY-SA 4.0 and Wikipedia CC BY-SA.
Questions about this entry
Short answers drawn from the linked project or page.
Use clm-train posttrain --source hf:boost-re8-hover-train-qwen3.6. The loader is clm_train.data.jev_boost, documented in docs/jev-boost.md.
Qwen/Qwen3.6-35B-A3B-FP8, embedded with vLLM pooling using the last token, L2-normalized to 2048 dimensions.
state_embeddings.pt, action_embeddings.pt, neg_embeddings.pt (zero-padded hard negatives), neg_counts.pt, single_tokens.npz for one-token actions and negatives, and metadata.json with one dict per row.
The train split holds 12,021 pairs and the validation split 1,290 pairs, each pair carrying one hard negative in this revision.
It follows the source's license: HoVer is CC BY-SA 4.0 and the Wikipedia text is CC BY-SA. The embedding vectors are a derivative of that text.
vLLM pooling of a one-token prompt is noise for this model, so those vectors are zero and single_tokens.npz points them at jackyk02/qwen3.6-35b-a3b-fp8-single-token-embeddings, applied on load by clm_train.data.stores.EmbeddingDir.
Items whose state shares a paragraph with a Jev suite item are dropped, near-copies of the gold option are removed from the negatives, and rows are capped at 20,000 by hash sample.
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
Hugging Face dataset of Jev-shaped retrieval and classification decisions built from MM-BRIGHT, holding Qwen3.6-35B-A3B-FP8 state, action, and hard-negative embeddings for post-training CLM heads.
Jev Models & Reimplementations#open-models#benchmark#evaluation
Community
A family of 0.8B-9B ophthalmic decision models that score candidate answers without autoregressive decoding, trained on RetiJev-Data: 78,158 clinical-decision and diagnosis questions and 50,074 guideline-grading questions. RetiJev-9B answers a case in 158 ms on one GPU.
Jev Models & ReimplementationsArticle#arxiv#open-models#benchmark
Community
Apache-2.0 Hugging Face repository for RSI-JEV-27b, a 27B multimodal decision model based on Qwen3.8-27B that answers supplied multiple-choice questions over images, returning a selected option and probabilities for all options in one forward pass.
Jev Models & Reimplementations#open-models#benchmark#calibration
Community
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
Hugging Face dataset that converts AVeriTeC claim-verification records into Jev Decision Index retrieval and classification requests, embedded with Qwen3.6-35B-A3B-FP8 into 2048-d vectors for post-training CLM heads.
Jev Models & Reimplementations#open-models#classification#benchmark
Community
Hugging Face dataset of Jev Decision Index-shaped state/action embeddings built from Re-DocRED. Holds 13,312 train and 1,517 validation pairs embedded with Qwen3.6-35B-A3B-FP8 at 2048 dimensions for post-training CLM heads.
Jev Models & Reimplementations#open-models#classification#benchmark
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.