artifact-open-science

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Artifact Open Science

成果物开放科学

Prepare research artifacts that reviewers and future researchers can actually run, inspect, and cite. Use this for artifact evaluation, reproducibility badges, camera-ready packages, and public release.

准备可供评审人员和后续研究人员实际运行、检查和引用的研究成果物。适用于成果物评估、可复现性认证、终稿提交包以及公开发布场景。

Read First

必读文档

references/artifact-open-science-policy.md

```
references/repository-contract.md
```
```
references/ethics-data-governance.md
```

references/artifact-open-science-policy.md

```
references/repository-contract.md
```
```
references/ethics-data-governance.md
```

Workflow

工作流程

Identify artifact type: code, dataset, benchmark, model, UI, pipeline, reproduction package, or supplementary material.
Define what claims the artifact supports.
Create or update
```
artifacts/artifact-checklist.md
```
.
Ensure install/run instructions are fresh, minimal, and tested.
Separate public release files from private data, credentials, caches, and large local outputs.
Add license, citation, expected runtime, hardware, seed/config, and known limitations.
Link artifact claims to
```
experiments/
```
,
```
repro_outputs/
```
,
```
sources/
```
, and
```
wiki/
```
.

确定成果物类型：代码、数据集、基准测试、模型、UI、流水线、复现包或补充材料。
明确该成果物支持的研究结论。
创建或更新
```
artifacts/artifact-checklist.md
```
文件。
确保安装/运行说明是最新、简洁且经过测试的。
将公开发布文件与私有数据、凭据、缓存和大型本地输出文件分离。
添加许可证、引用信息、预期运行时间、硬件要求、随机种子/配置以及已知限制。
将成果物支持的结论与
```
experiments/
```
、
```
repro_outputs/
```
、
```
sources/
```
和
```
wiki/
```
目录关联。

Minimum Artifact Standard

成果物最低标准

one-command smoke test or clear manual path
environment file or dependency list
data access/provenance instructions
expected outputs
result comparison against paper claims
ethical and license constraints

一键式冒烟测试或清晰的手动操作步骤
环境配置文件或依赖项列表
数据访问/来源说明
预期输出结果
与论文结论的结果对比
伦理和许可证约束

Do Not

禁止事项

Release sensitive or license-restricted data without governance notes.
Ship a notebook-only artifact when a script/CLI is expected.
Claim reproducibility if only exploratory outputs are available.

在未提供管理说明的情况下发布敏感或受许可证限制的数据。
在预期使用脚本/CLI时，仅交付笔记本形式的成果物。
仅提供探索性输出时，声称成果具有可复现性。