Polar codes are a novel class of capacity-achieving error-correcting codes. List decoding techniques are needed in order to implement near-maximum likelihood decoding of polar codes and their generalizations, such as polar codes with CRC and polar subcodes.  However, the complexity of the Tal-Vardy list decoding algorithm with large list size is too high for a practical implementation.  It can be reduced by employing sequential decoding techniques.  In this talk, we show that by carefully designing a score function, it is possible to obtain a decoding algorithm with better performance and substantially lower complexity compared to belief-propagation decoding of state-of-the-art LDPC codes. The proposed approach enables several further complexity reduction tricks.