Page limits without typographic crimes

Your paper is 10.4 pages. The limit is 10. What happens in the next hour determines whether you fix the paper or commit a formatting violation that gets it desk-rejected.

Squeeze the content, not the format

Work this list in order. Each step is invisible to format checkers because it changes the paper, not the template.

1. Tighten prose first. Half a page of overflow is usually a wordy related-work section and three redundant transition sentences. Cutting words improves the paper. Cutting whitespace does not.
2. Hunt widows. A paragraph whose last line holds one or two words wastes a full line. Rewriting one sentence in that paragraph reclaims it.
3. Resize figures honestly. Many figures have huge internal margins. Trim the whitespace in the figure itself, or drop a two-column figure to one column if it stays legible.
4. Slim tables. booktabs tables with no vertical rules are more compact and look better, see beautiful tables, and abbreviated column headers buy real width.
5. Use vector figures. PDF plots scale down cleanly, so you can shrink them further than pixelated PNGs before legibility dies. Fuzzy raster figures have their own problems, see fuzzy fonts in PDFs.
6. Move material to an appendix or supplement, if the venue allows it. Proofs, extra ablations, and hyperparameter tables are classic overflow candidates. Check whether the appendix counts against the limit, many venues exclude it, some do not.

The crimes

These change the template instead of the paper. Program committees run automated format checkers precisely because people try them:

Crime	How it gets caught
\vspace{-8pt} sprinkled everywhere	Visibly cramped spacing, diffs against the class defaults
\usepackage[margin=...]{geometry}	Automated margin measurement
Shrinking \baselinestretch or font size	Line counts per column, font size probes
\small or \footnotesize on body text or captions	Font size probes, reviewer eyeballs
Negative \itemsep, crushed lists	Cramped look, checker heuristics
Renaming the class file with edits inside	Checksum against the official class

Why venues actually check

A page limit is a fairness contract: every author gets the same space, and every reviewer commits to reading a bounded amount. A paper that cheats its way to 10 pages is really an 11-page paper competing against honest 10-page papers. That is why violations are treated as an integrity issue, not a style nit: many venues desk-reject without review, and there is no appeal that says “but the content was good”.

The honest last resort

If you have tightened the prose, slimmed the floats, moved proofs to the appendix, and you are still over: cut a result. Painful, but a focused 10-page paper reviews better than a crammed one anyway. Reviewers can feel density, and choosing a venue with a longer format is always an option for the next submission.