On top of the excellent summaries so far, here are a few things I try to do every game if I'm not at war.
Early commerce tactics - run binary research (0%/100%) to avoid rounding losses and to exploit library multipliers.
binary science is indeed very useful, but not for the reasons you mention. rounding errors, for the vast majority of the game, are insignificant, and binary science only benefits more from multipliers in that it is more beneficial to build up gold right before a big research multiplier comes online and then switching to full science afterwards, as more of your total commerce over a turn-span will get multiplied by the new building. (and vice versa, when weath multipliers like banks come online). otherwise, there is no net benefit. assuming no changes to your empire, running at 50% science slider for +0 gold/turn yields the same net beakers, regardless of what academies, libraries, universities, whatever you have, as running at 0% science slider for +100 gold turn for 1 turn and then running at 100% science for -100 gold/turn for 1 turn.
what binary science provides is 2 things:
1.) binary science lowers tech costs in some situations. that is to say, if you don't think you'll the first to a tech, then its better to delay your beaker investment into that tech, as the total beaker cost of any tech decreases for every other civ that knows it. this is very helpful on the higher difficulty levels, where you may be first to only a few techs, if any, all game.
2.) binary science gives you more strategic flexibility in your research path. lets say you're faced with a difficult choice on which tech to go after next. for example: do i try to win the free artist from music, or do i go straight for bureaucracy? if you were to know that you'll be first to music, you think this will benefit you more than 10 turns without bureaucracy, but, facing against other strong techers, you're not sure if you can win it. if you research halfway there, at, say, 50% slider, and then get beat by huayna capac, then you just wasted quite a few turns pursuing a tech that won't have a lot of value. he got the artist, will beat you to the wonder (or reduce your total failgold by finishing it fast, if that's what you wanted), and the cherry on top is that tech doesn't even have full trade value anymore.
if, on the other hand, you build up your gold reserves first, and then run your slider at full rate, you'd discover that the music path was a bad idea right before you actually invested anything into it. you can now switch to researching bureaucracy, and get it at the same date you would have if you never even considered music in the first place.
binary science helps in this regard even when the tech doesn't have a huge prize attached to being first. for example, many leaders won't trade away techs unless at least 2 other civs know it. so, you might be able to get the tech you want in a trade instead of researching it yourself, and then research something else instead!