CommunityJev Models & Reimplementations
Lavoir
A decision model that scores its own clarifying questions by value of information: at AUC .799 against an oracle .797, it asks only when asking is expected to change the answer, then decides or hands off.
- Published by
- Community
- Author
- MoganAI
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Returns calibrated option probabilities and a value-of-information estimate for every question it could still ask.
- Question-asking curve AUC .799 against an oracle that knows the exact routing rules at .797.
- Asks in 6% of real SGD conversations; where it does not ask, its decision is right 96% of the time.
- Ahead of Laya's English checkpoint on 7 of 12 single-turn benchmarks, level on spam and phishing.
- One forward pass, 31 ms median on a GH200, with no text generation.
Watch out
Weights are CC-BY-NC-4.0, so commercial use is not covered, and the evaluation numbers are the card's own. The value-of-information head is bounded by the decision's Gini impurity, so a confident model structurally will not ask; a Turkish sibling trained on MoganBERT-TR is published separately as Lavoir-TR.
Questions about this entry
Short answers drawn from the linked project or page.
For every candidate question it returns how much the probability of the right answer is expected to rise if that question is asked, bounded by the decision's Gini impurity as (1 - sum of p squared) times sigmoid(h).
Ask the question with the highest value of information while it exceeds ask_threshold; then decide the top-probability option, or hand off when 1 - max p exceeds handoff_threshold, which defaults to 0.3.
It asks in 6% of real SGD conversations, and where it does not ask its decision is right 96% of the time; its question-asking curve scores AUC .799 against an oracle's .797.
It is ahead of Laya's English checkpoint on 7 of 12 single-turn benchmarks and level on spam and phishing, and without slots the input is exactly Laya's format, so it behaves as a Laya-style classifier.
CC-BY-NC-4.0 weights, with the question-asking training and test data in moganai/lavoir-dialogues; a Turkish version trained on MoganBERT-TR is published as Lavoir-TR.
It makes one forward pass with no text generation, at a median of 31 ms on a GH200; the decision head scores each option marker and applies a softmax with a fitted temperature.
A string, a dict of named fields, or a list of conversation turns using 'user' for the customer and 'system' for the assistant's questions.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
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 NewsHacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsMore Jev resources like this
The v3 Jev-Style decision model at 2B in GGUF: F16 3.78 GB and Q8_0 2.01 GB match PyTorch FP32 on 100% and 99.7% of parity rows, 73.6% on JevBench v1.4.1 public items.
Jev Models & Reimplementations#community#open-models#local
Community
Mithril's open, Jev-shaped typed-decision model on DeBERTa-v3-large: choice up to 255 options, score and noul from one forward pass, 0.854 in-domain with ECE 0.022, trained on public gold labels only.
Jev Models & Reimplementations#community#open-models#local
Community
TokenRhythm's 4B prefill-only decision model built on NeoHorse-1-4B: given a state and application-defined questions it returns Choice, Noul, and Score probabilities without generating text, reporting a 77.70 six-group aggregate under Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
An unofficial cross-encoder fine-tune of ModernBERT-ja-310m for Jev Local: it reads a question-plus-state and one candidate per sequence and hits 92.62% on the JGLUE JNLI test, with CC BY-SA 4.0 weights.
Jev Models & Reimplementations#community#open-models#local
Community
The 322M mmBERT-base build of Laya for 100+ languages: state plus typed questions return typed answers with probabilities in one forward pass, with up to 8,192 tokens for long documents.
Jev Models & Reimplementations#community#open-models#local
Community
ggmlc-compiled GGUFs of the Laya typed-decisions specialist on ModernBERT-large: F16, Q8_0, and UD_Q4_K_M files for invoice, security, customer-service, and harness presets, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.
Jev Models & Reimplementations#open-models#training#local
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
Read article
From the Jev community
Posts from builders shipping with Jev right now.