References are not nouns. What does this mean? It means that the following is wrong: “In [1], Jack and Jill went up a hill.” Instead, you can say, “Mother Goose reports that Jack and Jill went up a hill [1].”

Now that Halloween is over, go on a which hunt. In general, everywhere you are using “which,” you should be using “that.” The particular rule is that “which” prefaces a clause that could be removed from the sentence without changing the meaning; if you need the clause to understand the sentence, then you want “that.” A corollary is that if the phrase is preceded by a comma, you probably want “which.”

An example:

The car that is in the garage is broken.

The 1982 Toyota Corolla, which was decomissioned last year, was named Beauregard.

Note the example in the text above. Why is “that” correct instead of “which?”

Keep “only” close to its clause. If you use the word “only,” push it as close as possible to the clause to which it applies.

If you have one subject and two predicates, do not separate the predicates with a comma.

Laundry lists of references are often worthless. Avoid things like, “Many people have investigated caching [1][2][3][4][5][6].” Instead, tell us something about what the references actually say. “While many people have studied caching, only one study shows that it is a fundamentally flawed idea [1]. Several others indicate that it is the best thing since sliced bread [2][3][4], and a few authors actually seem to provide an accurate evaluation [5][6].” (Hint: this means that you actually have to have read the related work.)

Avoid “very.” I believe it was Mark Twain who said, “Substitute d–n every time you’re inclined to write ‘very’; your editor will delete it and the writing will be just as it should be.”

While you are at it, avoid “attempts”. You do things in research papers. You might be more or less successful at doing those things, but don’t weasel-word it.

Avoid passive voice. It is significantly easier and more enjoyable to read technical prose written in the active voice.

Your paper should not read as a murder mystery. Tell the reader in the beginning, that’s in the abstract, what the significant results of the research are. There is nothing more frustrating than reading a paper wondering the whole time what you’re going to see.

Use fewer and less appropriately. Fewer implies that you are describing a discrete quantity while less implies a continuous one. So, you might have fewer people or things, but you might have less inclination or less clutter in your room.

Use ‘like’ and ‘such as’ correctly. Use like only when A) the example you give is not exact or B) if the thing following the like can be properly embellished with a verb (typically do or does). In all other cases, use ‘such as.’ Examples:

Case A: Honeysuckle bushes, like flowers, have a strong aroma. (The bushes are not actually flowers, but we are using flowers as an example.)

Case B: 'That flower smells like my perfume.' -- the complete sentence could be 'That flower smells like my perfume does.'

Proper use of 'such as' : 'Graph-processing systems such as GraphChi, Ligra, ....'

Impact is probably the most over-used term in technical papers today. People use it as a verb, they use it as a noun. I use it to describe wisdom teeth (impacted). Just avoid it. Use affect(s) and effect(s) (details).
Another phrase I despise is "seeks to." Disks seek to places, but research papers don't; research projects don't; and you shouldn't either.
I also find that the word "aim" sneaks into papers these days with alarming regularity. We aim to do this; the system aims to do that; people aim to do something else. Make claims, don't waffle. We do! Our system does! The other folks do something different. I'm pretty sure that you're not thinking, "Well, we aimed to do X, but failed." If you are saying that you aimed for something, you are also claiming that you hit it, so just say that.
"And so" should almost always be replaced by "so," and so just do it that way.
Figures illustrate things; people visualize things. Do not say, "Figure 1 visualizes ..." A figure can illustrate something and sometimes it can present a visualization of something, but the figure does not visualize anything.
Ah yes, another phrase to avoid, "We/I argue ..." Don't argue with your reader. You might point out or note things. You might try to persuade the reader. But mostly you want the prose to do that for you without telling the reader that you're trying to pull a fast one over on him or her.
Run a spell checker. It won't catch everything, but it should catch the things that will embarrass you.
Assume that your readers are going to believe you and agree with you instead of trying to convince them of something before you've even presented it. Say, "The sun rises in the East, because the earth rotates in an easternly direction." instead of, "Because the earth rotates in an easternly direction, the sun rises in the East." In the first case, if the reader agrees with your statement s/he will skim over your explanation and simply nod. In the latter, you're providing an explanation out of context and the reader is going to think about it and try to decide if s/he agrees with you before going on to read the point that you actually care about. And the point that you care about is your statement, not the explanation.
Composed of versus Comprises: A big thing is composed of little things. A bunch of little things comprise a big thing. Get this right.
I hate the word incentivize. Incent means the same things and has half the syllables.
I hate the phrase "This is because ..." It typically means that the previous sentence should be combined with this one, using ", because ..."
Use 'may', 'can', and 'might' correctly. In general, 'may' implies permissions and is rarely what you mean in technical writing.
Focus on the four C's: Writing should be Concise, Crisp, Clear, and (grammatically) Correct.
[Tips on avoiding gender-specific writing](http://herodios.com/pronouns.html)
[Strunk and White (which should probably be read annually)](http://www.bartleby.com/141/)