Disclosed is a method by which a terminal decodes a codeword in a wireless communication system. Specifically, the method may comprise: receiving a plurality of codewords; and decoding the plurality codewords on the basis of successive interference cancellation (SIC). In particular, the SIC may be performed on the basis of a decoding policy for decoding the plurality of codewords. In particular, the decoding policy may be determined by a neural network trained on the basis of a state and a reward related to the plurality of codewords.