Jev directory
CommunityJev Benchmarks & Evaluations
JevGAG Adaptive Temporal Granularity Figures
Hugging Face dataset of 28 figure files (13.5 MB) supporting the study Jev-Routed Temporal Granularity with Recovery and Hysteresis for Causal Video Generation, including cost-quality Pareto plots, Jev routing traces, controller ablations, and qualitative frame grids.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- 28 rows of manuscript figures and frame grids, 13.5 MB total, licensed CC-BY-4.0.
- Covers cost-quality Pareto plots on Tesla T4 and A100-SXM4-80GB, signal AUROC, and controller ablation.
- Routing compares frame-wise c1-1, chunk-3 c3-3, and a near-zero-cost hold mode under a two-threshold controller.
- Includes an architecture diagram, qualitative frame grids, and contact sheets of decoded frames.
- Companion datasets hold the result videos and the audited results E1-E9.
Questions about this entry
Short answers drawn from the linked project or page.
It holds publication figures and frame grids for the study Jev-Routed Temporal Granularity with Recovery and Hysteresis for Causal Video Generation: manuscript figures, an architecture diagram, qualitative frame grids, individual qualitative start and end frames, and contact sheets of decoded frames.
The dataset has 28 rows in a single train split, listed in the under-1K size category, with a total file size of 13.5 MB. Files are image and document modalities in imagefolder format, auto-converted to Parquet for the viewer.
The dataset card states a CC-BY-4.0 license, so the figures can be reused with attribution. The card also lists a citation entry for the JevGAG preprint by MD Jisan Hossen.
The study audits ran on a Tesla T4 and an A100-SXM4-80GB. The figures include cost-quality Pareto plots for both, hardware scaling and break-even, and A100/T4 online policies.
The routing controller is driven by the Jev System One oracle. A two-threshold controller with hysteresis and forced recovery routes between frame-wise c1-1, chunk-3 c3-3, and a near-zero-cost hold mode.
They are published as separate companion datasets: one holding the result videos and one holding the audited results E1-E9. The card also points to code repositories, project pages, a Kaggle source, and two Zenodo preprints.
It is maintained by Vision Intelligence (MD Jisan Hossen, NUIST). It contains figures and frame grids only, so anyone needing the decoded videos or the full audited result tables should use the companion datasets instead.
More Jev resources like this
Benchmark comparison of an untrained Qwen3.5-9B against hosted Jev across decision, tool-calling and calibration benchmarks. Reports accuracy, latency and ECE on matched tasks, plus a local adapter that serves Qwen through Jev's typed systemone API.
Jev Benchmarks & Evaluations#evaluation#calibration#benchmark
Case study evaluating Jev 1.13.0 on time-series decision tasks, including classification, event type prediction, oil temperature direction, and stock movement, converting continuous forecasts into discrete decisions and scoring accuracy and AUC.
Jev Benchmarks & EvaluationsArticle#arxiv#evaluation#benchmark
Community
arXiv paper introducing PrimaCons, a four-axis benchmark measuring whether Jev gives consistent decisions when the same semantic state is expressed as different typed query primitives, plus the Cross-Primitive Residual Calibrator (CPRC) proposed to reduce that inconsistency.
Jev Benchmarks & EvaluationsArticle#arxiv#evaluation#calibration
Community
arXiv paper scoring candidate pages by the citation share they will receive, comparing pretrained rerankers, Jev, eight open decision models and distilled students on GEO-Bench queries answered by ten engines. Releases GEO Citation Ranker 4B, the benchmark, all scorer predictions and code.
Jev Benchmarks & EvaluationsArticle#arxiv#evaluation#benchmark
Community
arXiv study evaluating Jev as a decision reader of truncated LLM reasoning across seven reasoners and five multiple-choice benchmarks, comparing accuracy and estimated cost against forced answering at budgets from 500 to 16,000 tokens.
Jev Benchmarks & EvaluationsArticle#arxiv#evaluation#benchmark
Community
arXiv study comparing Jev 1.13 with Clef, Microsoft-Decision-1, Laya, GPT-6 Luna and GPT-6 Sol on 172 rheumatology exam questions and 70 diagnostic vignettes, reporting accuracy, calibration and language effects.
Jev Benchmarks & EvaluationsArticle#arxiv#evaluation#calibration
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.