Updates from the NeurIPS 2025 Position Paper Track
Barbara Engelhardt and Katherine Gorman, Position Paper Track Chairs 2025
As the Position Track Chairs for NeurIPS 2025, we are so excited to bring this new section of the conference to life. Inspired by the fantastic work done at ICML in 2024, the Position Paper Track is intended to spur conversation in the community about pressing topics. We’ve built on the amazing work that Kiri Wagstaff and her team have done but we’re hoping to cut our own path and make the Position Paper tracks at our two conferences distinct and unique.
We’re providing updates on our track as it recently hit some big milestones: as of May 22nd, we closed submissions, conducted our initial review, and set out desk rejects. We received 657 submissions to the track, of which 140 (21%) received a desk rejection.
Our criteria for desk rejections reinforced standards set in the Call for Papers, which required that papers “simulate (productive, civil) discussion on timely topics that need our community’s input”. When doing our initial review, if we had trouble understanding an argument, if the paper was not formatted correctly, or if the title of the paper did not speak to the argument, these did not meet our criteria. We prioritized papers that were clear and whose positions were easily accessible. As a new track for NeurIPS, we sought to ensure these criteria were fairly applied and consistent. To this end, we implemented a two-pass review. All papers were manually screened, sometimes with 3 PCs and APCs reviewing the paper, and no AI was used to do any of the screening. In a second pass, all desk rejected papers were reviewed a second time, to ensure confidence in our decisions. This desk rejection phase is necessarily to maintain the quality and selectivity of this track, and to protect the time and energy of our reviewers and area chairs.
We also found an interesting trend in submissions: many submissions were positioned as “semi-technical papers”, which made interesting technical arguments in conjunction with their positions. The general argument of these papers is that the community should use their new or published technical approach for a broad swath of problems, with benchmarks as evidence. Following our guidelines, we desk rejected these papers, as they are best suited for the technical track, and not the position track, because they are focused in scope and raise questions better addressed by technical reviewers. These semi-technical papers were not a great fit for the goals of our track. Alternatively these might be in a workshop focused on the methods they discuss.
We also stumbled a little in the communication of our process, for which we apologize. If you received a desk reject, you probably got only a few notes in OpenReview as to why. We’re truly sorry for our terse communication, and we’ve been communicating with the authors (one by one!) via email to help clarify. In most cases, we’ve had to maintain our position on the desk rejection, but there have been a few papers that have passed on to the review process after (pretty intense) internal debate. Please be ensured that these short comments are not reflective of the rigor of our process, which considered each desk rejected paper in two independent passes to ensure the rejection was founded.
As the criteria for position papers are more subjective than the community is used to in general, we’re currently considering the possibility of building a public appeals process. We’re hoping this will help to further our core goal of stimulating conversation and debate. If you have ideas for what would make such a process successful, we would love to hear them!
We hope NeurIPS Position Papers will create an entirely new space for discussion of pressing issues and topics in our community, and we are so grateful for the effort and energy that everyone has put into helping us move its development forward! Overall, as a new track for NeurIPS, we’re aiming to be as transparent as possible, towards the aim of collecting feedback and improving the track in future years. We look forward to making this first iteration of the Position Track a success with the help of the entire community, and we are excited to hear about the discussions our track stimulates at the NeurIPS Conference and beyond!