You read faster the moment you stop trying to get through a paper from page one in a single sitting. This is not an argument for skimming. It is an argument for reading the same paper three times, each with a different purpose.
Pass 1 — read in order to discard
Read only the title, abstract, section headings, figure captions, and conclusion. Five minutes.
There is exactly one question to answer: is this relevant to my question?
If it is not, close it. The purpose of the first pass is not comprehension — it is deciding whether the rest of your time is worth spending here. The output of this pass is not the number of papers you read. It is the number you threw away.
Pass 2 — read in order to get the claim
Read the body, but skip the proofs, the experimental details, and the appendices. About an hour.
When you finish, you should be able to explain three things to someone else:
- what the paper claims
- what it is claiming to be better than
- under what conditions the claim is said to hold
Look at the figures before the prose. The experimental design is usually written on the axes of a chart more clearly than it is written in the text.
Pass 3 — read in order to check reproducibility
Do this only when your own work will depend on the paper. Several hours.
Write out the assumptions one at a time and read while tracing how you would implement it yourself.
The purpose of this pass is to find what the authors did not write down. Initial values, preprocessing, stopping criteria — once you know where the unwritten parts are, you can judge whether it is safe to build on top of this paper.
Decide in advance how far you will go
The three-pass method works not because the passes are separate, but because you decide each time whether to go on to the next one. You do not need to read every paper three times. Most stop at the first pass, and that is the correct outcome.