Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
103 lines (59 sloc) 7.81 KB

Tips for submitting a proposal or paper to PAPIs — and a few rules

This document completes the guidelines for the PAPIs Call for Proposals.

How to craft a good talk proposal

Attendees have different levels of technical expertise, and also different levels of knowledge of Machine Learning (from beginner to advanced). Start by identifying your main audience and the level of technicality of your talk. Your presentation should either directly help them, or inspire/inform them about something they don’t already know. The core value to our attendees of what you’re presenting should be clearly stated in your proposal. What will they be able to do after they see your presentation that they can’t do now?

We believe that great presentations should be practical and focused. For instance, instead of presenting a portfolio of things you have done or a list of problems you have solved, it’s better to show one specific, unique thing in enough detail. We also highly encourage to show live demos, when possible; this can be very useful to make your presentation more concrete and engaging.

Some examples of popular presentations at previous PAPIs conferences:

How to submit your proposal

Our proposal form has two fields that attendees will see on the program (Title and Abstract), and two fields that only reviewers will see (Details and Pitch). Please use tags to help reviewers identify the type of proposal you’re submitting.

It is possible to submit several proposals through our CfP application. However, we have a policy of not having more than 2 talks from speakers from the same company, in an effort to promote different approaches to ML applications at the conference.

Title, Session Format, Track, Abstract & Bio

These are what attendees will see in the program. The title and abstract should be compelling and to-the-point. Tell a story. Why should attendees come to your presentation and what will they get out of it?

Please identify the track which is best suited to your proposal (see our CfP guidelines for track descriptions), while avoiding to select the "general" track. If you think your talk could also fit in another track, please explain in the Details field.

The default session format is 20' Presentation. However, if you are submitting to the Tools track, you must choose between 10' Demo or 30' Tutorial.

Your bio won’t be seen by reviewers in the 1st round of reviews but we will use it when advertising your presentation if it is accepted. Please include your current position and organization.


Only reviewers will see this. Please take care here to refrain from identifying who you are since our first round of review is blind, and we appreciate your efforts to respect that as much as possible.

It's ok to use the abstract as a teaser of your presentation, but please use the Details field to go into more depth about what you’ll cover and to be as specific as possible on the content of your presentation. We invite you to...

  • include an outline (with rough time estimates)
  • reveal the "secret sauce" of your presentation
  • explain any twists you’ll include that may not be evident in the title or abstract
  • let us know if you will also be sharing code or an API (and links to them, if any).


Only reviewers will see this. Pitch is a good place to tell reviewers why PAPIs needs this presentation, and why you’re the right person to give it at PAPIs. How will your presentation help the program, or fill a specific need? Why are you excited about this topic?


Please use tags to identify your target audience (ML beginners, ML practitioners, ML experts) and the level of technicality of your talk (non-technical, somewhat technical, very technical).

Please tag your proposal with demo if you're planning a include a demo, and explain how long you plan the demo to be in the Details field.

Please tag your proposal with paper if you're also submitting a paper (read more below).

Industrial experience reports, technical, review and research papers

Industry reports or papers are not required to get your proposal accepted, but they're highly appreciated. We will publish accepted papers in Proceedings of Machine Learning Research, after the conference. Some examples:

Publishing a paper in a peer-reviewed journal requires some effort but it is another great way to contribute to your reputation and that of your company. Previous authors include teams at Microsoft, Uber Engineering, Upwork, Dataiku, BigML... Papers are very valuable for the community as they provide more details on what you did, for others to learn from.

Submitting a paper (or extended abstract) with your proposal

Please prepare one of the following anonymized documents in PDF format to include to your talk proposal:

  • an extended abstract of 2 pages or more (references and comparisons to related work should be included, but details of implementation can be omitted)
  • a short paper of 4 pages
  • a long paper of up to 8 pages.

Please share this as a download link in the Details field. (You can create a link by uploading your paper to a cloud service such as Dropbox or Google Drive. We will keep all links private.)

Preparing your camera-ready paper

Authors should use LaTeX to prepare their camera-ready papers (if you're not familiar with LaTeX, feel free to contact us once your initial submission has been accepted, so we can help). Please use the sample file jmlrwcp-sample.tex as a starting point for formatting your article; we will provide a starting page number and volume number for PMLR, once your proposal has been accepted.

You should prepare your submission as a ZIP archive containing...

  • your paper as a single LaTeX .tex document
  • any accompanying images in an images directory
  • the compiled paper as a PDF document of 4-8 pages in length (not including bibliography)
  • a completed release form.

Please then share this archive as a download link in the Comments section of the CfP app.

Compiling the LaTeX source into PDF

The provided LaTeX file uses the jmlr class with the pmlr option: \documentclass[pmlr]{jmlr}. This class is defined in the jmlr.cls style file. You can find some more details about that class in jmlr.pdf, and full details at CTAN.

Any questions?

If you have questions or concerns about anything you see here, please contact us.


These tips were partly inspired from the railsconf CfP guidelines.