Appendices in conference papers have become a shadow main text, and that defeats the purpose of a page limit. When a paper at NeurIPS or ICML requires a 25-page appendix to understand its core claims, the system has quietly abandoned the principle that a conference paper should stand on its own. The original intent was clear: the main pages contain the essential contribution, and the appendix holds secondary material. That line has eroded.
For researchers, this shift creates a practical burden. If reviewers routinely demand additional experiments that cannot fit into eight pages, the appendix stops being optional and becomes a requirement for acceptance. You are then writing two papers, one for the conference, one for the archive, and the shorter version no longer tells the full story. This does not serve rigor; it rewards the ability to pack more content into supplementary sections, often at the expense of clarity. A reader who skips the appendix misses critical evidence, yet the main paper is supposed to be the complete narrative.
The argument that twenty-five pages belong in a journal is reasonable. Journals are designed for depth, for extended experimental sections, for the kind of thoroughness that a conference format cannot accommodate. Conferences trade depth for speed and community feedback. When the appendix becomes the real paper, we lose that trade-off. We also risk normalizing a culture where page limits are treated as a hurdle to be circumvented rather than a constraint that sharpens focus.
Our view is simple: if your contribution requires twenty-five pages to validate, submit it to a journal. If you need a conference venue, respect the page limit as a design constraint, not a suggestion. Appendices should support the main paper, not replace it. The community should enforce this not through stricter rules alone, but by rewarding concise, self-contained contributions in the review process. That is the standard worth returning to.