The main thing is going to be practice. I recommend trying this. Write down two answers for the inference questions. You should get to the point where you can narrow it down to two answers. If you are consistently narrowing it down to two answers, the next step is going to be understanding the minute detials that makes one right and one wrong.

One strategy that may be helpful with Inference questions:

Consider the statements given in the stimulus to be not only facts, but constraints. Anything other than those constraints is in the realm of possibility when you try to create a hypothetical that would invalidate the proposed conclusion.

So, as an example, say the stimulus said:

All savings accounts bear interest, and some interest bearing accounts are tax-free.

Which of the following conclusions can be drawn?

A) Some savings accounts are tax-free
B) Some savings accounts are not tax-free
C) Some savings accounts are interest-bearing

If you look at the given information as constraints, you know that you can come up with any hypothetical as long as:

Savings accounts are all interest-bearing

and

Some interest bearing accounts are tax free

So, to attack answer choice A, come up with a hypothetical that shows that it's not true: Savings accounts and IRA accounts bear interest, but only IRA accounts are tax free. This hypothetical is consistent with the premises, but invalidates choice A, so A is not necessarily true.

To attack B, you can create another hypothetical: It could be that all savings accounts are tax free, and some IRA accounts are tax free. This is consistent with the idea that some interest-bearing accounts are tax free, so it's allowed by the stimulus, and it invalidates B.

C cannot be disproven. If you try to come up with the hypothetical that "certain savings accounts do not bear interest", the premise that "all savings accounts bear interest" won't let you - you're stuck with the given information, so C must be true.

Often times, Inference questions lend themselves really well to process of elimination. Because the correct answer "must be true" given the premises, then any choice that lends itself to being "not necessarily true" is incorrect. If you treat the given information as a way to check the hypotheticals that you draw, you can employ that elimination strategy effectively.