CommunityJev Models & Reimplementations
Decider Chat Gemma-4-31B
An untrained readout: stock Gemma-4-31B weights plus a decider config that reads option letters at the answer slot, scored #2 of 70 on Decision Index v0.2.1 at 57.33 against Jev 1.13.0's 57.91.
- Published by
- Community
- Author
- Mapika
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- No fine-tune at all: identical Gemma-4-31B-it weights, a fitted temperature function, and a readout that reads one answer slot.
- Decision Index v0.2.1 score 57.33 with ECE 0.047 and a 108.5 ms median latency, measured by the index rather than self-reported.
- A replay of 300 stored rows through decider.serve matched 1,412 of 1,412 answers, median probability difference 0.0001.
- Temperature scales with option count: T(n) = max(0.05, 10.124 − 1.633 ln n).
Watch out
The behaviour is exactly the base model's — the readout adds only a fixed answer slot and a temperature, and the card says the knowledge, reasoning, and failure modes are unchanged. The checkpoint is 62.5 GB in bf16, so the vLLM path is the practical one, and temperature_by_options needs decider-ai 1.8.0 or newer.
Questions about this entry
Short answers drawn from the linked project or page.
No training at all. The repository holds unchanged Gemma-4-31B-it weights plus a decider_config.json; all behaviour comes from the readout, and the card says its knowledge, reasoning, and failure modes are the base model's.
On Decision Index v0.2.1 it scores 57.33 to Jev 1.13.0's 57.91, placing second of 70 with 0.047 ECE. The card notes the index measured it rather than the authors, and its serve path exposes POST /v1/systemone.
It needs decider-ai>=1.8.0: Decider("Mapika/decider-chat-gemma4-31b").system_one(...) runs it directly, and DECIDER_MODEL=Mapika/decider-chat-gemma4-31b uvicorn decider.serve:app exposes POST /v1/systemone. For the 62.5 GB checkpoint, decider.serve_vllm serves the same readout.
Temperature scales with option count, T(n) = max(0.05, 10.124 - 1.633 ln n) for a question with n options. Gemma-4 instruct models are overconfident with few options and underconfident with many, so one global temperature cannot calibrate them.
No. The repository holds the unchanged Gemma-4-31B-it weights plus a decider_config.json, and the readout adds only a fixed answer slot and a fitted temperature, so the knowledge, reasoning, and failure modes are the base model's.
The Gemma-4-31B-it weights are Apache-2.0 as published, and the readout code in github.com/Mapika/decider is Apache-2.0 too. There is no training, so the licence covers only the unchanged base weights and the config.
A replay of 300 stored index rows through decider.serve picked the same answer on 1,412 of 1,412 answers, with a median probability difference of 0.0001. The index measured it on one RTX PRO 6000 at a 108.5 ms median.
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 NewsRelated terms
Glossary definitions related to this entry.
More Jev resources like this
The fine-tuned Laya checkpoint for four typed-decision workflows: 0.766 accuracy on 2,000 held-out decisions, above Jev 1.13.0's published 0.727 and the teacher self-agreement ceiling it trained past.
Jev Models & Reimplementations#community#open-models#local
Community
Supersonic Labs' 144.3M-parameter decision model on mmBERT-small: choice, score, and noul answers from supplied options in one pass, 73.15% on 2,000 typed decisions, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Liquid AI's 3.12B decision model built on LFM2.5-VL: state plus typed questions in one forward pass with zero output tokens, 48.57 on Decision Index 0.2.1 and 8 ms per decision on an RTX 4090.
Jev Models & Reimplementations#community#open-models#local
Community
Nace.AI's 8B decision model on NVIDIA's Efficient-DLM diffusion backbone: a 256-dim pointer head scores every option in one pass, behind the same POST /v1/systemone API as its hosted service.
Jev Models & Reimplementations#community#open-models#local
Community
The vLLM Semantic Router project's 29.37B decision model on Qwen3.8-27B: a probability per supplied option with zero generated tokens, 74.0 on JevArena and 56.5 on Jev Decision Index 0.2.1.
Jev Models & Reimplementations#community#open-models#local
Community
A 4B scorer rebuilt from Qwen3-4B into a branching cross-encoder: one branch per option, no position bias, up to 256 options, and calibrated probabilities with no text to parse.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
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
Base decision checkpoints sit near chance on purpose. The typed-decisions loop, the corpora behind the students, and the temperature fit that turns probabilities into thresholds.
Jev Practices & Patterns#fine-tuning#calibration#open-models
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
From the Jev community
Posts from builders shipping with Jev right now.