An engine cycle where intake valves only open in the last part of intake cycle?
 

Would this work?
Has this been tried before?

If during the intake of the 4-stroke cycle
intake valves remain closed, say, during the first half of the piston downward travel.
Some vacuum MIGHT form during the piston downward travel.
Then during the last half, the intake valves open, sucking air fuel mixture all of sudden.

I said MIGHT as some warm exhaust gas might be intentionally left in the cylinder, which will expand, compensating some of the vacuum created by the piston moving down.

I believe this will reduce part-throttle pumping losses.

We call that retarding the intake phase.
Its normally used to boost horsepower not MPGs.
Early opening and early closing has always been used to increase fuel economy.

seems like it would drop efficiency, since work would be required to generate the vacuum in the cylinder. the amount of residual gasses would have to be pretty significant to prevent a vacuum from being formed.

for the "instant suck" to happen, there would have to be a vacuum in the cylinder, the larger the vacuum, the more significant the "sucking".

Late and/or early intake valve closure has been used before for efficiency improvement. It's typically called Miller cycling or Atkitson cycling (depending on who you ask). The Toyota Prius engine uses this. The effect is an expansion ratio which is greater than the compression ratio. What you're suggesting wouldn't change the effective compression ratio and I'm not sure as how it would reduce pumping losses. Remeber that lower cylinder pressure on the intake stroke is going to increase pumping losses.

Considering just the idealised cycle on a PV diagram there will be unrecoverable pumping work done during the part of the "intake" stroke where the valve remains closed.

This is different to early or late inlet valve closing where the pressure remains trapped in the cylinder and the work done in generating it can be recovered.

Instead of trying to get around pumping losses why not switch to a cycle that benefits from filling the cylinders with as much air as possible?

I'll take the opportunity to mention one of my favorite things, the Scuderi Split-Cycle

Scuderi engine - Wikipedia, the free encyclopedia
http://upload.wikimedia.org/wikipedi...ne_-_Cycle.gif

Making an assumption about what you are implying :):

All else being equal a CI engine (idealized as constant pressure combustion) is less efficient than a SI engine (idealized as constant volume combustion).

Of course all is not equal and there remains potential in trying to minimise the losses in SI engines.