This contribution reports the test results of CE6-4.3 on RQT-like transform sub-block splitting. With the proposed RQT scheme, the transform block can be only split once (maximum transform depth is 1), and splitting type is implicitly derived based on the block shape. For inter prediction residuals, transform types are also implicitly derived for each sub-block based on its relative location inside a coding block. It is reportedly shown that, on top of VTM-3.0, average -0.30%, -0.42% and -0.45% luma BD-Rate is achieved under All Intra (AI), Random Access (RA) and Low Delay B (LDB) configurations, respectively.