I find it a little annoying that in the paper[0] they show various graphs of megabytes <i>saved</i> in the paper, but no actual <i>size</i> of the binaries that these policies are applied to, as far as I can tell.<p>So when they say the inline policies end up saving 20 MiB on the training data, and then only a few megabyte on a different binary not in the training data, I lack the context to really judge what that says. Is the other binary much smaller? The same size? What if it's <i>bigger</i> and therefore hides a much smaller relative size savings?<p>At the very end of the paper do they mention one binary size: namely that they save about 3 MB on the Chrome on Android binary, which is 213.32 MB after implementing the policy. A solid 1%, probably makes an enormous difference at Google Scale, especially for their main Android browser, so I hope it's obvious that I'm not trying to diminish the achievement of these people. But I find the other benchmarks kind of hard to interpret.<p>[0] <a href="https://arxiv.org/pdf/2403.19462" rel="nofollow">https://arxiv.org/pdf/2403.19462</a>