Without your CSV file to test, it is impossible to reproduce.
See How to Ask The Community for Help .
I suspect that if you are using Japanese characters, they cannot be represented by a standard ASCII CSV file and you must use a CSV with UTF-8 encoding. I have had problems with importing UTF-8 CSV files (exported from Excel) into ISE for network devices and network device groups so I suspect this is the problem.
I recommend contacting TAC and potentially filing a bug on the lack of UTF-8 support for CSVs in ISE.