I'm not aware of any general rule of thumb. More practically, it comes down to tuning it for the circuit and design constraints at hand.
Of course, the most efficient dead time is zero, but the closer you get to zero the greater the likelihood of shoot-through. Typically shoot-through is extremely undesirable, so we take a conservative approach. How conservative is up to the particular design - how well controlled are the parameters of gate capacitance, gate charge current, gate discharge resistance and threshold voltage? How much variation over the operating temperature range must be tolerated? These are the kind of parameters which will determine how short you can set the dead-time.
The other side of the coin is the relative gain in efficiency you can achieve. With such a high switching frequency, small gains in dead time may well be significant. But it'll be a case of diminishing returns. Look at the percentage of time the circuit spends transitioning between the hard-on states. Is it worth trying to make it any shorter? Is doing so likely to cause shoot-through at some corner of the operating domain?