Study: NLP research is migrating from ACL to general ML venues
TL;DR
- In the post-LLM era, established NLP authors lost 19.2pp of share at flagship *ACL main-conference tracks while gaining 14.8pp in Findings tracks.
- General ML venues rose 8.6pp among established authors, even after adjusting for parallel growth across both fields.
- Among debut authors with three or more first-author NLP papers, those mostly at *ACL fell from 84% in 2019 to 74% in 2024.
A paper from David Jurgens on arXiv is worth a slow read if you care about where NLP research actually lives now. Using publication data from 2010 to 2026, Jurgens argues that the disciplinary center of gravity of NLP is shifting away from its flagship conferences and toward general machine learning venues.
The numbers he lands on are specific. Comparing pre- and post-LLM eras, established authors lost 19.2 percentage points of their share at flagship *ACL main-conference tracks, while gaining 14.8pp in the newer Findings tracks and 8.6pp at general ML venues, even after adjusting for parallel growth in the fields. Among newer authors who debut with at least three first-author NLP-topic papers, the share whose work appears mostly at *ACL venues fell from 84% in 2019 to 74% in 2024, while the share appearing mostly at general ML venues rose from 5% to 21%.
Why that matters if you are not tracking conference politics: the flagship ACL series has been the primary place practitioners look for NLP methodology and benchmarks. If a growing share of the work that used to land there is now going to general ML venues instead, the practical effect is that keeping up with NLP means reading across more places, and the incentives that shape which problems get worked on start being set by a broader audience. Jurgens uses causal inference techniques to estimate that general ML venues confer a significant citation premium, which he reports as one of the factors influencing venue selection.
The honest caveats are worth naming. This is a preprint, and the abstract does not break down which subfields of NLP are migrating fastest, or which specific ML venues are absorbing the shift. The claim about the citation premium is a modelled estimate from causal inference, not a settled measurement. Take the specifics as reported, not as final.
If the migration continues, the venues doing the real steering of NLP research a few years from now may not be the ones with NLP in the name.
Shared on Bluesky by 2 AI experts
Originally reported by arxiv.org
Read the original article →Original headline: [2607.02416] The Future of NLP may not be at NLP Conferences: Scholarly Migration Patterns in Natural Language Processing