A scientific paper consists of a constellation of artifacts that extend beyond the document itself: software, hardware, evaluation data and documentation, raw survey results, mechanized proofs, models, test suites, benchmarks, and so on. In some cases, the quality of these artifacts is as important as that of the document itself. Last year, 84% of accepted OSDI papers participated in the artifact evaluation process. Based on the success of the OSDI '21 Artifact Evaluation last year, this year USENIX ATC will run an optional artifact evaluation process for the first time, in combination with OSDI.
The artifact evaluation process will consider the availability and functionality of artifacts associated with their corresponding papers, along with the reproducibility of the paper's key results and claims with these artifacts. Artifact evaluation is single-blind. Artifacts will be held in confidence by the evaluation committee.
All (conditionally) accepted USENIX ATC papers are encouraged to participate in artifact evaluation. Because the time between paper acceptance and artifact submission is short, we strongly encourage authors to start preparing their artifacts for evaluation while their papers are still under consideration by the USENIX ATC Program Committee. See the Submitting an Artifact section for details on the submission process.
Questions about the process can be directed to firstname.lastname@example.org.
- Notification for paper authors: Friday, April 29, 2022
- Artifact registration deadline: Tuesday, May 10, 2022, 8:59 pm PDT
- Artifact submission deadline: Thursday, May 12, 2022, 8:59 pm PDT
- Kick-the-tires response period: Thursday, May 19–Friday, May 20, 2022
- Artifact decisions announced: Tuesday, June 7, 2022
- USENIX ATC final papers deadline: Thursday, June 9, 2022
Note: For an artifact to be considered, at least one contact author for the submission must be reachable via email and respond to questions in a timely manner during the kick-the-tires period.
Artifact Evaluation Committee Co-Chairs
Anuj Kalia, Microsoft
Neeraja J. Yadwadkar, University of Texas at Austin
Chengyu Zhang, ETH Zurich
Artifact Evaluation Committee
Vojtech Aschenbrenner, EPFL
Vlad-Andrei Bădoiu, University Politehnica of Bucharest
Haoran Cai, Huawei Technologies Co.
Lianjie Cao, Hewlett Packard Enterprise
Siyuan Chai, University of Illinois at Urbana–Champaign
Stefanos Chaliasos, Imperial College London
Gohar Irfan Chaudhry, Microsoft Research
Hongzheng Chen, Cornell University
Pengfei Chen, Sun Yat-sen University
Shawn Chen, Carnegie Mellon University
Weikeng Chen, DZK Foundation and University of California, Berkeley
Weiteng Chen, University of California, Riverside
Weihao Cui, Shanghai Jiao Tong University
Dong Dai, University of North Carolina at Charlotte
Sowmya Dharanipragada, Cornell University
Boyuan Feng, University of California, Santa Barbara
Yebo Feng, University of Oregon
Joshua Fried, MIT CSAIL
Swapnil Gandhi, Microsoft Research India
Bin Gao, National University of Singapore
Wei Gao, Nanyang Technological University
Seyed Armin Vakil Ghahani, University of Michigan
Ayush Goel, University of Michigan
Cheng Gong, Nankai University
Junzhi Gong, Harvard University
Rong Gu, Nanjing University
Yunfei Gu, Shanghai Jiao Tong University
Shengjian Guo, Baidu Security
Dongjie He, University of New South Wales
Chen-Yu Ho, King Abdullah University of Science and Technology
Xiaofeng Hou, The Hong Kong University of Science and Technology
Qinghao Hu, Nanyang Technological University
Tianhao Huang, Massachusetts Institute of Technology
Xuanteng Huang, Sun Yat-sen University
Md Shahriar Iqbal, University of South Carolina
Abdullah Al Raqibul Islam, University of North Carolina at Charlotte
Vivek Jain, University of California, Riverside
Jae-Won Jang, Virginia Tech
Jianyu Jiang, The University of Hong Kong
Zhifeng Jiang, Hong Kong University of Science and Technology
Saisha Kamat, University of North Carolina at Charlotte
Arun Kp, IIT Kanpur
Saurabh Kumar, Indian Institute of Technology
Fadhil Kurnia, University of Massachusetts Amherst
Kahina Lazri, Orange Labs
Hugo Lefeuvre, The University of Manchester
Changlong Li, East China Normal University
Mingyu Li, Shanghai Jiao Tong University
Tao Li, Nankai University
Xiaolu Li, Huazhong University of Science and Technology
Yuan Li, Tsinghua University
Zongjie Li, Hong Kong University of Science and Technology
Yu Liang, City University of Hong Kong
Bo Lin, National University of Defense Technology
Jiacheng Liu, Shanghai Jiao Tong University
Jiawei Liu, University of Illinois at Urbana–Champaign
Jing Liu, University of Wisconsin—Madison
Zhibo Liu, The Hong Kong University of Science and Technology
Jie Lu, The Institute of Computing Technology of the Chinese Academy of Sciences
Mulong Luo, Cornell University
Minghua Ma, Microsoft Research
Suyash Mahar, University of California, San Diego
Shaghayegh Mardani, The University of Texas at Austin
Ines Messadi, Technische Universität Braunschweig
Zeyu Mi, Shanghai Jiao Tong University
Sumit Kumar Monga, Virginia Tech
Pu Pang, Shanghai Jiao Tong University
Qiangyu Pei, Huazhong University of Science and Technology
Chao Peng, ByteDance
Bin Qian, Newcastle University
Haoran Qiu, University of Illinois at Urbana–Champaign
Md Ashfaqur Rahaman, University of Utah
Nathan Rutherford, Royal Holloway, University of London
Mahsa Saeidi, Oregon State University
Bolarinwa Olayemi Saheed, University of Pécs
Amit Samanta, University of Utah
Divyanshu Saxena, The University of Texas at Austin
Carlos Segarra, Imperial College London
Vishwanath Seshagiri, Emory University
Sara Mahdizadeh Shahri, University of Michigan
Ridwan Shariffdeen, National University of Singapore
Rajath Shashidhara, University of Washington
Bogdan Alexandru Stoica, University of Chicago
Foteini Strati, ETH Zurich
Cesar Stuardo, University of Chicago
Qiang Su, City University of Hong Kong
Chendong Wang, University of Wisconsin—Madison
Jingbo Wang, University of Southern California
Pengcheng Wang, Purdue University
Qiang Wang, Tencent
Qiuping Wang, The Chinese University of Hong Kong
Shangwen Wang, National University of Defense Technology
Yongfeng Wang, Sun Yat-sen University
Yuke Wang, University of California, Santa Barbara
Zheng Wang, Nanyang Technological University
Dongwei Xiao, Hong Kong University of Science and Technology
Minhui Xie, Tsinghua University
Zhiqiang Xie, Max Planck Institute for Software Systems (MPI-SWS)
Le Xu, The University of Texas at Austin
Yunhong Xu, Texas A&M University
Yuqi Xue, University of Illinois at Urbana–Champaign
Zhifei Yang, EPFL
Peisen Yao, The Hong Kong University of Science and Technology
Zhisheng Ye, Peking University
Edgardo Barsallo Yi, Purdue University
Boxi Yu, The Chinese University of Hong Kong
Guangba Yu, Sun Yat-sen University
Liangcheng Yu, University of Pennsylvania
Yuanyuan Yuan, The Hong Kong University of Science and Technology
Ji Zhang, Huazhong University of Science and Technology
Shulai Zhang, Shanghai Jiao Tong University
Xie Zhang, Sun Yat-sen University
Zhen Zhang, Johns Hopkins University
Zhe Zhao, ShanghaiTech University
Chen Zou, University of Chicago
Daming Zou, ETH Zurich
Benefits and Goals
The dissemination of artifacts benefits our science and engineering as a whole. Their availability encourages replicability and reproducibility and enables authors to build on top of each others' work. It can also help more unambiguously resolve questions about cases not considered by the original authors. It also confers direct and indirect benefits to the authors themselves.
The goal of artifact evaluation is to incentivize authors to invest in their broader scientific community by producing artifacts that illustrate their claims, enable others to validate those claims, and accelerate future scientific progress by providing a platform for others to start from. A paper with artifacts that have passed the artifact evaluation process is recognized in two ways: first by badges that appear on the paper's first page, and second by an appendix that details the artifacts.
Eventually, the assessment of a paper's accompanying artifacts may guide the decision-making about papers: that is, the Artifact Evaluation Committee (AEC) would inform and advise the Program Committee (PC). For now, artifact evaluation will begin only after paper acceptance decisions have already been made. Artifact evaluation is optional, although we hope all papers will participate.
Each paper sets up certain expectations and claims of its artifacts based on its content. The AEC will read the paper and then judge whether the artifacts match those criteria. Thus, the AEC's decision will be that the artifacts do or do not "conform to the expectations set by the paper." Ultimately, the AEC expects that high-quality artifacts will be:
- consistent with the paper
- as complete as possible
- documented well
- easy to reuse, facilitating further research
Authors will be invited to submit their artifacts after their papers have been (conditionally) accepted for publication at USENIX ATC. Because the time between paper acceptance and artifact submission is short, the AEC chairs encourage authors to start preparing their artifacts for evaluation while their papers are still under consideration by the USENIX ATC Program Committee. See the guidelines for packaging artifacts later in this document.
At artifact-submission time, a submitter will choose the criteria by which their artifacts will be evaluated. The criteria correspond to three separate badges that can be awarded to a paper. An artifact can meet the criteria of one, two, or all three of the following badges:
- Artifacts Available: To earn this badge, the AEC must judge that the artifacts associated with the paper have been made available for retrieval, permanently and publicly. We encourage authors to use Zenodo, which is a publicly-funded long-term storage platform that also assigns a DOI for your artifact. Other valid hosting options include institutional repositories and third-party digital repositories (e.g., FigShare, Dryad, Software Heritage, GitHub, or GitLab—not personal webpages. Other than making the artifacts available, this badge does not mandate any further requirements on functionality, correctness, or documentation.
- Artifacts Functional: To earn this badge, the AEC must judge that the artifacts conform to the expectations set by the paper in terms of functionality, usability, and relevance. In short, do the artifacts work and are they useful for producing outcomes associated with the paper? The AEC will consider three aspects of the artifacts in particular.
- Documentation: are the artifacts sufficiently documented to enable them to be exercised by readers of the paper?
- Completeness: do the submitted artifacts include all of the key components described in the paper?
- Exercisability: do the submitted artifacts include the scripts and data needed to run the experiments described in the paper, and can the software be successfully executed?
- Results Reproduced: To earn this badge, the AEC must judge that they can use the submitted artifacts to obtain the main results presented in the paper. In short, is it possible for the AEC to independently repeat the experiments and obtain results that support the claims made by the paper? The goal of this effort is not to reproduce the results exactly, but instead to generate results independently within an allowed tolerance such that the main claims of the paper are validated.
After the artifact submission deadline, members of the AEC will download each artifact package, read the accepted paper, install the artifacts (where relevant), and finally evaluate the artifacts. AEC members may communicate with artifact authors—through HotCRP to maintain the evaluators' anonymity—to resolve minor issues and ask clarifying questions. Authors must respond to messages from the AEC in a timely manner for their artifacts to be effectively considered.
The AEC will complete its evaluation and notify authors of the outcomes. Authors can use the time between notification and the final paper deadline to incorporate feedback and artifact details into the final versions of their papers. This is intended to allow authors to include the feedback from the AEC, at their option.
When the AEC judges that an artifact meets the criteria for one or more of the badges listed above, those badges will appear on the final version of the associated paper. In addition, the authors of the paper will be encouraged to add an Artifact Appendix of up to two pages to their publication. The goal of the appendix is to describe and document the artifact in a standard format. The template for the appendix is available here.
The AEC will try to accept any kind of digital artifact that authors wish to submit: software, data sets, survey results, test suites, mechanized proofs, etc. Paper proofs will not be accepted, because the AEC lacks the time and often the expertise to carefully review paper proofs. Physical objects, e.g., computer hardware, cannot be accepted due to the difficulty of making the objects available to members of the AEC. (If your artifact requires special hardware, consider if/how you can make it available to evaluators online.)
The submission of an artifact does not give the AEC permission to make its content public. AEC members may not publicize any part of your artifact during or after completing evaluation, nor may they retain any part of it after evaluation. Thus, you are free to include models, data files, proprietary binaries, etc., in your artifact. Participating in artifact evaluation does not require you to later publish your artifacts (although it is encouraged).
Some artifacts may attempt to perform malicious or destructive operations by design. These cases should be boldly and explicitly flagged in detail in the README so the AEC can take appropriate precautions before installing and running these artifacts. Please contact email@example.com if you believe that your artifacts fall into this category.
Review and Anonymity
Artifact evaluation is "single blind." The identities of artifact authors will be known to members of the AEC, but authors will not know which members of the AEC have reviewed their artifacts.
To maintain the anonymity of artifact evaluators, the authors of artifacts should not embed any analytics or other tracking in the websites for their artifacts for the duration of the artifact-evaluation period. If you cannot control this, do not access this data. This is important to maintain the confidentiality of the evaluators. In cases where tracing is unavoidable, authors should notify the AEC chairs in advance so that AEC members can take adequate safeguards.
Submitting an Artifact
Registration and Submission
Submitting the artifacts associated with your accepted USENIX ATC paper is a two-step process.
- Registration: By the artifact registration deadline, submit the abstract and PDF of your accepted USENIX ATC paper, as well as topics, conflicts, and any "optional bidding instructions" for potential evaluators via the artifact submission site.
- Submission: By the artifact submission deadline, provide a stable URL or (if that is not possible) upload an archive of your artifacts. If the URL is access-protected, provide the credentials needed to access it. Select the criteria/badges that the AEC should consider while evaluating your artifacts. You will not be able to change the URL, archive, or badge selections after the artifact submission deadline. Finally, for your artifact to be considered, check the "ready for review" box before the submission deadline.
The AEC recommends that you create a single web page at a stable URL that contains your artifact package. The AEC may contact you with questions about your artifacts if your submitted materials are unclear.
The review process is structured in two phases:
- Kick-the-tires: During this phase, reviewers will check for any obvious problems that prevent the artifact from being fully reviewed. Such problems include invalid download links, broken virtual machine images, missing dependencies, or failures when applying the artifact to a "Hello world"-sized example. Authors can respond to issues and provide an updated version of their artifact during a kick-the-tires response period.
- Full evaluation: After the kick-the-tires phase, reviewers will fully evaluate the artifact.
The goal of the Artifact Evaluation Committee is to judge whether the artifacts that you submit conform to the expectations set by your paper in the context of the criteria associated with the badges you have selected. The effort that you put into packaging your artifacts has a direct impact on the committee's ability to make well-informed decisions. Please package your artifacts with care to make it as straightforward and easy as possible for the AEC to understand and evaluate their quality.
A complete artifact package must contain:
- the accepted version of your USENIX ATC paper
- the artifact itself
- README instructions
README instructions: Your artifact package must include an obvious "README" that describes your artifact and provides a road map for evaluation. The README must consist of two sections. A "Getting Started Instructions" section should help reviewers check the basic functionality of the artifact within a short time frame (e.g., within 30 minutes). Such instructions could, for example, be on how to build a system and apply it to a "Hello world"-sized example. The purpose of this section is to allow reviewers to detect obvious problems during the kick-the-tires phase (e.g., a broken virtual machine image). A "Detailed Instructions" section should provide suitable instructions and documentation to fully evaluate the artifact.
Artifact claims: Importantly, make your claims about your artifacts concrete. This is especially important if you think that these claims differ from the expectations set up by your paper. The AEC is still going to evaluate your artifacts relative to your paper, but your explanation can help to set expectations up front, especially in cases that might frustrate the evaluators without prior notice. For example, tell the AEC about difficulties they might encounter in using the artifact, or its maturity relative to the content of the paper.
Artifact format: Authors should consider one of the following methods to package the software components of their artifacts (although the AEC is open to other reasonable formats as well):
- Source code: If your artifact has few dependencies and can be installed easily on several operating systems, you may submit source code and build scripts. However, if your artifact has a long list of dependencies, please use one of the other formats below.
- Virtual machine/container: A virtual machine or Docker image containing the software application already set up with the right toolchain and intended runtime environment. For example:
- For raw data, the VM would contain the data and the scripts used to analyze it.
- For a mobile phone application, the VM would have a phone emulator installed.
- For mechanized proofs, the VM would contain the right version of the relevant theorem prover. We recommend using a format that is easy for AEC members to work with, such as OVF or Docker images. An AWS EC2 instance is also possible.
- Binary installer: Indicate exactly which platform and other run-time dependencies your artifact requires.
- Live instance on the web: Ensure that it is available for the duration of the artifact evaluation process.
- Internet-accessible hardware: If your artifact requires special hardware (e.g., SGX or another trusted execution environment), or if your artifact is actually a piece of hardware, please make sure that AEC members can somehow access the device. VPN-based access to the device might be an option.
- Screencast: A detailed screencast of the tool along with the results, especially if one of the following special cases applies:
- The artifact needs proprietary/commercial software or proprietary data that is not easily available or cannot be distributed to the committee.
- The artifact requires significant computation resources (e.g., more than 24 hours of execution time to produce the results) or requires huge data sets.
- The artifact requires specific hardware or software that is not generally available in a typical lab and where no access can be provided in a reasonable way.
As previously described, in all cases, artifacts must be provided in a manner that is appropriate for single-blind review by members of the AEC (i.e., anonymous reviewers).
There are several sources of good advice about preparing artifacts for evaluation. These two are particularly noteworthy:
- HOWTO for AEC Submitters, by Dan Barowy, Charlie Curtsinger, Emma Tosch, John Vilk, and Emery Berger
- Artifact Evaluation: Tips for Authors, by Rohan Padhye
If you have any questions about how best to package your artifact, contact firstname.lastname@example.org.
The AE process at USENIX ATC '22 was inspired by multiple other conferences, such as USENIX Security, SOSP, OSDI, and several SIGPLAN conferences. See artifact-eval.org for the origins of the AE process, and sysartifacts.github.io for the previous AE processes held in systems.