Read in passes rather than straight through from start to finish.
First pass: title, abstract, figures and conclusion to decide if it matters to you.
Second pass: method and results, skipping the proofs.
Third pass: work through the details closely enough to reimplement it.
Most papers deserve only the first pass, and that is fine.
A paper is not a novel and reading it start to finish is the slowest way through.
Read in passes, and stop as soon as the paper stops earning your time.
The first pass is the title, abstract, figures and conclusion.
Five minutes tells you what was done and whether it matters to you.
Most papers deserve only that, and abandoning them is the correct outcome.
The second pass covers the method and the results, skipping proofs and derivations.
Look hard at the figures and tables, since the real claim is usually there.
Check what the baselines were and whether they were tuned as carefully as the method.
A weak baseline is the most common way an improvement is manufactured.
The third pass works through the details closely enough to reimplement it.
That is where you notice the assumption that quietly does the work.
Read the ablations, since they say which part of the idea actually matters.
Read the limitations section, which is often the most honest paragraph in the paper.
Look at when it was published, because the field moves fast enough for two years to matter.
Reading the code alongside resolves more confusion than rereading the text.
One question on this card. Every option explains itself once you pick, right or wrong.
11 concepts in this module, then a graded test with every answer explained.