# Emmanuel's list of paper writing and research tips This is my list of tips about things that sometimes get on my reviewer nerves (thus the prolific use of "!") even though they're kind of minor. Well, not always. I usually add to this list after a round of reviewing/writing. Take everything with a grain of salt. Last update: 2024-05-21 - **Don't use too many conjuctive adverbs!** If your paragraph has more than one "Thus", "Therefore", or "Furthermore", rethink it. You rarely actually need those words because English is read left-to-right; order your sentences the right way and you'll rarely need a "Thus". - **Get really good at making plots!** I swear the next paper I see with labels for ants or a tensorboard screenshot will... get my eyes to roll. More seriously though, you should be making at least a plot a day; and I don't mean looking at loss curves, I mean summary plots. Did you try different L2 $\lambda$ values today? Make a plot with $\lambda$ as the $x$ axis and $y$ as whatever metric makes sense. - **Plot fonts matter**, if your text is 12pt, your labels and legends should be at least 9pt, ideally 10pt. - **Colors, markers, and fills matter**, make strategic use of sequential or uniform colormaps, make your plot as concise and informative as possible. - **Try different things**, like slogans, there are always multiple ways to present something. - **Cache, cache, cache**, save intermediate computations, if you need to re-plot something 3 hours before the paper deadline, it better be fast & accessible. - **Don't work on weekends!** Your body and mind will thank you. If the pressure is too high, try to at least not work every other weekend. Doesn't have to be the weekend. - **Get decently good at LaTeX!** Proper use of `\citet{..}`, `\citep{..}`, `\citep[page 42][]{newton1692}` is silly but so often wrong, please get it right I beg you. Make pretty algorithms, figure captions, you'll look so _professional_. (Also knowing how to quickly type out (and read) TeX math is :100:) - Learn `tikz`? This is definitely a bonus (and a steep hill) but if you want to make sexy diagrams don't look further than `tikz`. - **Ask a(n English) friend!** English is not my native tongue and I would not be where I am now without having learned from others. If you don't have such friends, many chatbots will now be happy to correct your English for free. It's worth the time investment. _[[Conversely, as a reviewer, please be generous towards obviously non-English authors]]_ - **Use the present tense** I don't think I could explain why, but as a stylistic choice it just feels much neater. - **Start writing early!** This will not only diminish stress before a deadline but also as you write your draft you'll notice (with $p=0.9$ in my experience) that some cool important plot/experiment is missing. - **Have a lead author!** On overleaf & co, have someone whose job it is to accept/reject suggested changes. This is typically just the first author but doesn't have to be. I've seen too many drafts where the document ends up being a rainbow of suggested edits, making it hard to distiguish edits which require discussion from minor edits. I'll end by saying that I also mess up all of these things from time to time, unfortunately. No one is perfect, but aiming higher is good :)