Jev directory
CommunityJev Models & Reimplementations
UltraBERT-Jev 380M Softmax Yes
A 380M-parameter UltraBERT encoder fine-tuned as a Jev-style decision model: it reads a state, a question, and a list of options and picks one option in a single forward pass. Apache-2.0 weights hosted on Hugging Face, scored by softmax over each option's " yes" logit.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- Scores each option from masked-token logits at a mask placed in front of it, with no text generation.
- Mean accuracy over 13 datasets: 71.1% on the dev split and 68.7% on the test split.
- Trained for one epoch at batch 32 and learning rate 1.5e-5 on the 13-dataset Reason+Long mixture, with Nimble added.
- Load in fp32: bf16 weights changed accuracy by up to 8 points on long MuSiQue documents.
- 429,966,336 F32 parameters, apache-2.0 license, tested with transformers 5.20.2 and torch 2.9.
Questions about this entry
Short answers drawn from the linked project or page.
Use AutoTokenizer and AutoModelForMaskedLM from the repo Athekunal/ultrabert-jev-380m-softmax-yes, loading with dtype=torch.float32. It was tested with transformers 5.20.2 and torch 2.9, which include the UltraBERT model class. Do not pass trust_remote_code=True, because the base repo's remote code names the weights differently and would leave them unloaded.
It reads a state, a question, and a list of options and picks one option in a single forward pass without generating text. Option k is scored from the masked-token logits at its mask, and the recipe is a softmax over each option's " yes" logit.
The masked-token logits are large, around 200, and bf16 weights changed accuracy by up to 8 points on long documents in MuSiQue during testing. All reported numbers use fp32 weights.
Accuracy is the mean over 13 datasets: 71.1% on the dev split and 68.7% on the test split.
It is UltraBERT-380M fully fine-tuned for one epoch at batch 32 and learning rate 1.5e-5 on the 13-dataset Reason+Long mixture of LLM-as-Jev, with Nimble added.
The repository is licensed apache-2.0 and holds 429,966,336 F32 parameters in a single unsharded safetensors file of about 1.72 GB. It is not quantized and is not gated.
No. The model was trained on a specific prompt format using the tokens " yes" and " no", and plain fill-mask use will not give the reported results. Use the prompt shown in the model card exactly as written.
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
A 1.7B-parameter SocialJev preview model on Hugging Face, fine-tuned from Qwen3-1.7B for typed classification in the social sciences. Runs through Transformers with trust_remote_code or the social-jev package with a vLLM backend, returning a choice and probabilities per query.
Jev Models & Reimplementations#open-models#classification#python
Community
A 4B-parameter SocialJev preview model on Hugging Face, fine-tuned from Qwen3-4B for typed classification in the social sciences. Loads through Transformers with trust_remote_code or the social-jev package with a vLLM backend, returning a choice and probabilities per query.
Jev Models & Reimplementations#open-models#classification#python
Community
A Hugging Face checkpoint that fine-tunes UltraBERT-380M into a Jev-style decision model: it reads a state, a question and a list of options and picks one in a single forward pass. Reports 72.6% mean accuracy over 13 datasets on the test split.
Jev Models & Reimplementations#open-models#classification#local
Community
A cross-encoder set model for ROR set point matching, distilled from OpenAlex's Jev judgments. Given an affiliation string and its candidate ROR records, it gives each candidate the probability that the string names it.
Jev Models & Reimplementations#open-models#python#classification
Community
MLX precision variants of Jev-Style-2B-Decision-v4.2, a 1.88B-parameter text-only decision model for Apple silicon. Ships bf16 and 8-bit weights with a custom runtime that returns a probability for every supplied option.
Jev Models & Reimplementations#open-models#local#python
Community
LoRA adapter for Qwen3.5-4B that scores medical multiple-choice options in a single forward pass and returns a short JSON object with the chosen answer key and per-option probabilities, without explanations or chain-of-thought.
Jev Models & Reimplementations#open-models#python#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.