Lattice reduction (LR) techniques have been applied to linear equalizers to improve their performance. Seysen's algorithm (SA) is an LR method which has been shown to have better performance than the well-documented LLL algorithm in many cases. However, no matter which LR method is adopted, there still exists a performance gap between LR-aided and ML equalizers. To fill this gap, one could use soft detectors. In this paper, we first give a tree-search implementation of Seysen's method, which reduces the complexity of SA without sacrificing its performance. Based on this algorithm, an LR-aided soft detector is proposed to achieve better performance. The performance-complexity tradeoff of our proposed algorithms is also studied. Simulation results validate the effectiveness of our algorithms.