Jev directory
CommunityJev Models & Reimplementations
RSI-JEV-27b
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.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- Answers supplied multiple-choice questions over images, returning a chosen option and probabilities for all options in one forward pass.
- Public Decision Index 0.3, with overall micro accuracy of 81.4087% over a 12,210-request Vision panel.
- The repository holds all 18 base-weight shards, tokenizer and processor files, two unmerged LoRA adapters and the decision head.
- Recorded A800 per-request wall times: mean 777.36 ms, median 520.24 ms, P80 893.07 ms, P95 1,685.89 ms.
- Adapters and project code are Apache-2.0; benchmark image datasets are not included.
Questions about this entry
Short answers drawn from the linked project or page.
Follow inference/README.md in the Hugging Face repository for installation, a complete model download, Python examples and the official runner command. Load base/ plus bundle/ with both adapters active and unmerged using the supplied custom Engine.
No. The Hugging Face model repository contains all 18 base-weight shards, tokenizer and processor files, both LoRA adapters, the decision head and the exact evaluated inference source. The complete model is stored as base/ plus bundle/.
It is the local Public Decision Index for this model, using the official chance correction and weights rather than a simple mean of raw accuracies. The card states it is not the private or full leaderboard score, and no Text benchmark result is claimed.
The run used eight A800-SXM4-80GB GPUs, each running a separate batch-1 model instance. Per-request wall times were mean 777.36 ms, median 520.24 ms, P80 893.07 ms and P95 1,685.89 ms, including image processing and in-call compilation.
Model adapters and project code are released under Apache-2.0. The Qwen base and runtime dependencies retain their own licenses, and the vendored Decision Index SDK is MIT-licensed.
R-Bench-M has weight 0.5 and MMMU-Pro vision has weight 0.25, while the other nine benchmarks carry weight 1. Reported benchmarks include CV-Bench, BLINK, RealWorldQA, CharXiv, InfographicVQA, Mind2Web, Winoground, KIE and Moderation.
The base weights alone are not RSI-JEV-27b. The evaluated model keeps both LoRA adapters active and unmerged, so omitting the bundle or merging the adapters does not reproduce the published 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
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 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 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
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
Hugging Face dataset of Qwen3.6-35B-A3B-FP8 state/action embeddings for contrastive post-training on the Jev Decision Index's entailment and verification train splits: ANLI, ContractNLI, NLI4CT, HoVer and RAGTruth.
Jev Models & Reimplementations#open-models#calibration#research
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.