On the TOEFL you hear everything once. That single fact drives the whole skill: you cannot re-read audio, so listening is a prediction sport — you listen FOR things. Before the audio starts, read the question stems (not all options): they tell your brain what to listen for. During the audio, your job is not to understand every word; it is to build a rough map: what is this about, what are the 3-4 big points, where does the speaker change direction.
Main-idea questions ask what the talk is mainly about. The answer is almost always announced early ('Today we're going to look at…') and confirmed by what the speaker spends time on. Two traps: an answer that names a detail (true, but too small) and an answer that names the field, not this talk ('geology' vs 'evidence that continents move'). Test: if you had to text a friend the topic in five words, your text is close to the main idea.
Tested details are never random: they are examples, numbers, names, causes, steps in a process, and anything the speaker emphasizes, repeats or contrasts. When you hear 'for instance', 'in fact', 'the key point is', or a number — that is a detail worth a note. Untested details (dates mentioned in passing, decorative examples) exist to waste your attention.