Article
CommunityJev Models & Reimplementations
Flide: Accelerating Decision Inference with Exact State Forking
arXiv paper introducing Flide, an exact inference engine for decision models such as Jev. It decouples the logical tree of a decision request from physical execution and reports answers up to 9.5 times faster than vLLM and SGLang.
- Format
- Article
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- Flide is an exact inference engine targeting decision-model workloads such as Jev.
- It answers queries up to 9.5 times faster than vLLM and SGLang in the reported evaluations.
- ForkAttention navigates request dependencies via GPU-resident indices across standard attention and recurrent Gated DeltaNet layers.
- Every query branch is guaranteed to evaluate identically to an isolated forward pass in exact arithmetic.
- Branch topologies live in GPU-resident metadata, and agent histories can resume from checkpoints saved in earlier passes.
Questions about this entry
Short answers drawn from the linked project or page.
Existing generation-oriented inference engines handle decision workloads poorly, suffering from severe prefix recomputation overheads or distorted output probabilities. Flide decouples the logical tree structure of a decision request from its physical execution inside a single flattened token stream to avoid those costs.
The paper reports that Flide answers queries up to 9.5 times faster than vLLM and SGLang on public hybrid and decision models, while staying closest to a float32 reference in total variation distance among the compared engines.
ForkAttention is the mechanism Flide uses to navigate request dependencies through GPU-resident indices across both standard attention and recurrent Gated DeltaNet layers. It keeps branches addressable from metadata so a shared token stream can serve many query branches.
No. The paper states that Flide guarantees in exact arithmetic that every query branch evaluates identically to an isolated forward pass, so branch outputs are not approximated. It is also reported as closest to a float32 reference in total variation distance.
Yes. By expressing branch topologies in GPU-resident metadata, Flide lets expanding agent histories resume from checkpoints saved in earlier passes rather than recomputing the full prefix.
ForkAttention covers standard attention and recurrent Gated DeltaNet layers. A compact set of pre-recorded execution graphs lets the engine accommodate diverse request mixtures without a separate graph per topology.
It is a research paper describing an engine and its evaluations, not an installable product page. Readers looking for a released binary, license, or deployment guide will not find those details in the abstract material.
More Jev resources like this
Fine-tuning method that turns fix commits into comparison groups so an open-weight model scores code functions for vulnerability. An ensemble of five Qwen3.5-4B models reaches 0.814 AUROC on a 7,270-function benchmark, against 0.743 for Jev 1.13.
Jev Models & ReimplementationsArticle#arxiv#open-models#security
Community
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
arXiv paper proposing JevRec, which encodes recommendation items as variable-length semantic identifiers and adds a Jev-like decision modeling component with separate heads for length prediction and content generation, evaluated on three public datasets.
Jev Models & ReimplementationsArticle#arxiv#research#benchmark
Community
arXiv paper on a small candidate formalizer trained on offline teacher-generated lists that builds explicit, validated alternatives from task context for a frozen decision model, evaluated on CALVIN-D episodes and annotations.
Jev Models & ReimplementationsArticle#arxiv#research#benchmark
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
arXiv paper on Decision-Level Continual Learning from Typed Feedback, a method that keeps Jev-style accept and revoke signals in bounded channels above a frozen scorer, and introduces TCPA-Bench, which credits a correction only through its effect on later decisions.
Jev Models & ReimplementationsArticle#arxiv#research#evaluation
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.