You will simply change the interference pattern. By modifying the phase at different frequencies you will simply change the phase relationship between the two speakers and thus move the locations of any pressure nulls/nodes at that frequency. It would be pure luck as to whether this would 'reduce' the interference pattern; also depending on your definition of the reduce in this context.
With a single loudspeaker you can generally control the sound pressure at one location. Two speakers would give you control at two discrete points. N speakers, at N points. In a reverberant environment you will also need to identify the room response as well to achieve this.
The above rule can be broken, by using some kind of optimisation (least-squares as probably the simplest) to optimise the freqeuency response at N control locations toward a certain target (flat frequency response?), using M speakers, where M < N. You wouldn't get very far with 2 speakers though.
Read up on sound zones for some interesting background material on this and some rather extreme speaker configurations :)