AwesomeJev.github.io / site

# Awesome Jev

> Independent, bilingual collection of Jev and typed decision model research, models, datasets and tools. Not affiliated with TypeSafe AI.

## Main pages

- [Resource collection](https://awesomejev.github.io/): curated primary sources and verification labels.
- [Research guide](https://awesomejev.github.io/guide/): source-backed answers about Jev, typed decisions, probabilities and evaluation.
- [Papers](https://awesomejev.github.io/papers/): versioned preprints; findings remain author-reported.
- [Decision model directory](https://awesomejev.github.io/models/): independent releases, declared licenses and model-card evidence.
- [Developer tool directory](https://awesomejev.github.io/tools/): SDKs, workflow controls, inference runtimes and training tools, with documented limits.
- [New resources](https://awesomejev.github.io/discoveries/): new direct resources, related work and background research with evidence labels.
- [Research topics](https://awesomejev.github.io/topics/): focused, source-backed notes with direct answers and evaluation limits.
- [Jev moderation, stance and sentiment evaluation](https://awesomejev.github.io/topics/language-task-evaluation/): Read language-task results through policy labels, retrieval conditions, conversational context, extra inference and supervised postprocessing.
- [Jev calibration, abstention and confidence routing](https://awesomejev.github.io/topics/calibration-and-abstention/): Compare probability quality, missing-answer rejection, threshold transfer and cascade cost using six versioned studies and explicit evaluation controls.
- [Option-label bias in Jev-style decision models](https://awesomejev.github.io/topics/option-label-bias/): Separate option-label bias from order sensitivity using versioned research, conditional invariance claims and a practical diagnostic.
- [AudioJev: audio decisions, order calibration and release limits](https://awesomejev.github.io/topics/audiojev/): Read AudioJev’s paper, checkpoint and inference library together: candidate-order stability, seed-specific results, data overlap and separate licenses.
- [GraphDecisionBench: evaluating Jev on graph decisions](https://awesomejev.github.io/topics/graph-decision-benchmarks/): Understand graph structure, graph-text evidence and sequential construction as separate tasks, with versioned sources and reproduction limits.

## Chinese pages

- [资源索引](https://awesomejev.github.io/zh/): 精选一手来源与核查标签。
- [研究指南](https://awesomejev.github.io/zh/guide/): 有来源支撑的 Jev、类型化决策、概率与评测解答。
- [论文目录](https://awesomejev.github.io/zh/papers/): 按版本整理的预印本；结论仍为作者报告。
- [模型目录](https://awesomejev.github.io/zh/models/): 独立发布的模型、声明许可与模型卡证据。
- [工具目录](https://awesomejev.github.io/zh/tools/): SDK、工作流控制、推理运行时与训练工具，附文档中的限制。
- [新增资源](https://awesomejev.github.io/zh/discoveries/): 新增的直接相关资源、相关研究与背景研究，附证据标签。
- [研究专题](https://awesomejev.github.io/zh/topics/): 有来源支撑的专题笔记，附简明回答与评估局限。
- [Jev 审核、立场与情感任务评测](https://awesomejev.github.io/zh/topics/language-task-evaluation/): 从规则标签、检索条件、对话上下文、额外推理与监督后处理理解语言任务结果。
- [Jev 校准、拒答与置信度路由](https://awesomejev.github.io/zh/topics/calibration-and-abstention/): 结合六篇固定版本研究，比较概率质量、缺失答案拒答、阈值迁移和级联成本，并明确评测对照。
- [Jev 类决策模型的选项标签偏置](https://awesomejev.github.io/zh/topics/option-label-bias/): 区分选项标签偏置与顺序敏感性，结合固定版本研究、带条件的不变性主张和可操作的诊断。
- [AudioJev：音频决策、顺序校准与发布范围](https://awesomejev.github.io/zh/topics/audiojev/): 结合论文、检查点与推理库，理解候选顺序稳定性、单种子结果、数据重叠和不同许可。
- [GraphDecisionBench：怎样评估 Jev 的图决策能力](https://awesomejev.github.io/zh/topics/graph-decision-benchmarks/): 分别理解图结构、图文证据与序列构造任务，核对论文版本和公开材料的复现范围。

## Source formats

- [Guide, curated resources and research topics in Markdown](https://awesomejev.github.io/llms-full.txt)
- [Paper metadata](https://awesomejev.github.io/papers.json)
- [Model metadata](https://awesomejev.github.io/models.json)
- [Tool metadata and purposes](https://awesomejev.github.io/tools.json)
- [Expansion report](https://awesomejev.github.io/data/expansion-progress.json)
- [BibTeX](https://awesomejev.github.io/references.bib)
- [Repository](https://github.com/reacher-z/Awesome-Jev)
- [Verification methodology](https://github.com/reacher-z/Awesome-Jev/blob/main/docs/source-audit.md)

This optional file is a reading aid, not an indexing or ranking mechanism. Respect source versions and distinguish vendor claims, preprint findings, and independently reproduced evidence.