Jev directory
CommunityJev Models & Reimplementations
Jev Boost to6 Leyzer Home Embeddings (Qwen3.6-35B-A3B-FP8)
Hugging Face dataset of Jev Decision Index-shaped state, action, and hard-negative embeddings for the Leyzer home-appliance simulator, rendered with Qwen3.6-35B-A3B-FP8 at 2048 dimensions for post-training CLM heads.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- Ships float16 state, action, and hard-negative embeddings at 2048 dimensions for train and validation splits.
- Train split holds 3,478 pairs with up to 8 hard negatives each; validation holds 360 pairs.
- Embeddings come from Qwen/Qwen3.6-35B-A3B-FP8 via vLLM pooling, last token, L2-normalized.
- One-token actions and negatives are zeroed and resolved from a separate single-token embedding table on load.
- Total file size is 159 MB; the card reports 25 downloads in the last month.
Questions about this entry
Short answers drawn from the linked project or page.
State, action, and hard-negative embeddings for decisions shaped like the Jev Decision Index's Tool Use & Agentic Decisions requests, built from the Leyzer home-appliance simulator. Files include state_embeddings.pt, action_embeddings.pt, neg_embeddings.pt, neg_counts.pt, single_tokens.npz, and metadata.json per split.
Qwen/Qwen3.6-35B-A3B-FP8, using vLLM pooling with the last token, L2-normalized, at 2048 dimensions. The card lists max_model_len 16,384, state_head_tokens 2,048, renderer clm_train.render.Renderer, multimodal false, and negatives_k 8.
The card documents the command clm-train posttrain --source hf:boost-to6-leyzer-home-qwen3.6, feeding the clm_train.data.jev_boost path described in docs/jev-boost.md. The vectors are intended for post-training CLM heads.
The card states that vLLM's pooling of a one-token prompt is noise for this model, so one-token sequences are never embedded. They are zero in the vectors, and single_tokens.npz points them at jackyk02/qwen3.6-35b-a3b-fp8-single-token-embeddings, which clm_train.data.stores.EmbeddingDir applies on load.
At most 8 per row, with near-copies of the gold option dropped. neg_embeddings.pt is shaped [n, k, 2048] and zero-padded, with neg_counts.pt giving the real count per row. The train split has 3,478 pairs and validation has 360.
The dataset card lists the source's license, CC BY 4.0, and notes that the vectors are a derivative of that source text. The card also lists per-source licensing and libraries Datasets, pandas, and Polars.
cartesinus/leyzer-fedcsis at revision 30081334f95240cc1949efbe9a7d2d9b537db58a, train split: Leyzer's English thermostat, air-conditioner, and speaker commands, each with twins such as out-of-range temperature, already-set, unavailable, or clarify cases. Items whose state shares a paragraph with a Jev suite item are dropped, capped at 20,000 rows 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 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
Hugging Face dataset of Jev Decision Index-shaped state/action embeddings built from TimeQA. Holds 13,278 train and 1,460 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
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
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.
Jev Models & Reimplementations#classification#benchmark#open-models
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
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.