I think that the main approach to test a mapping algorithm is to simulate it with, e.g., HeMPS or NoCSim, and measure power consumption and performance. The main idea is that the different placement of application tasks affects these objectives. Therefore, by comparing results you can find an optimal solution.
I'm not aware of special benchmarks used for NoC mapping. Often multimedia examples have been used in the literature (e.g. H264), but not in a standardized way.
For the benchmarks to use, it depend on the tools that you are using, if you use GEM5 there are some benchmarks to use. In otherwise the most popular benchmarking used in the field of NoC mapping are the multimedia applications like MPEG4,