I'm working my way on a problem which requires me to store elements in fixed 4KB-sized pages on an SSD. Each page is requested once for some computation on its elements, and then the CPU requests for another different page from the SSD. I'm trying to ascertain whether any page contiguous-ness can be exploited here for smaller latency times. Any research papers or articles on this area would be very appreciated.
Short answer: no.
On an SSD, non-contiguous access (very slightly) increases the protocol and processing overhead - in theory, contiguous (or rather: longer) transfers could be slightly faster but you'd be hard pressed to actually measure that difference in real life, especially with NVMe.
That is caused in part by today's extremely fast processing capabilities but mostly due to the (almost) general overlap of I/O requests and data transfers using (deep) command queueing.
Note that I'm assuming a somewhat decent size of I/Os - most SSDs hit their peak (or even the interface's) with 4 KiB I/Os already.