Hm. I found it far more efficient to trim what you read instead of trying to read faster. At first, filtering at a paragraph level is something I found myself

Hm. I found it far more efficient to trim what you read instead of trying to read faster. At first, filtering at a paragraph level is something I found myself to do. Usually, I just read the first few words of the sentences in the paragraph (most of the time even just some sentences in the paragraph) until I mostly see what the purpose of the paragraph is about ("Ah, he is telling me about injective", for example). Once I know this, I either skip the entire paragraph or read it slower and deeper. Second, it is sometimes a good idea to rather start in the middle of something and read to the end, just backpatching from the first chapters whatever you still need. This works well if you are reading papers about things you are familiar with. You usually just don't need those 12 definitions again, and again, and again. I think, this is also what Knuth referred to as 'batch reading' of papers (and it being much more efficient).