A reliability metric is used for determining whether to prune a decoding hypothesis. For example, a reliability metric can be generated for each possible hypothesis generated during blind decoding operations. The reliability metric can then be used in a pruning process whereby a determination to prune a given hypothesis is based on whether the corresponding reliability metric is above or below a reliability metric threshold. In some aspects, the reliability metric is based on the correlation between the symbols of a hypothesis and re-encoded symbols that are based on the hypothesis, whereby the correlation is normalized using an estimated power parameter that is independent of the hypothesis. Through the use of the reliability metric, decoding may be achieved with a low probability of false passes (in the case of noise) and a low probability of missed detection (in the case of a real signal).