What gambits (if any) are actually an advantage, assuming perfect play in response?

From what I gather, a gambit is defined as risking or losing material, in an attempt to gain position and/or future advantage greater than what would be lost. 

My question is - are any gambits actually "correct" if you consider perfect play in response? And I don't really mean really late positions or extremely specific positions - more along the lines of early to mid game. That is, are there any "gambits" which don't rely on 'mistakes' in response?

The reason I ask is for coding a chess engine. If I can rule out things like gambits / moves where one side would be "losing" in the short term, I could really pare down the trees of moves that need to be analyzed.