The standard approach for Bayesian phylogenetics is the Markov chain Monte Carlo. Even though the asymptotic convergence is guaranteed for MCMC under weak assumptions, the computational cost is generally very expensive due to the curse of dimensionality and the complex multimodal distribution from the phylogenetic problem. Wang et.al  proposed a combinatorial sequential Monte Carlo (CSMC) method that can serve as a good alternative to MCMC in posterior inference over phylogenetic trees. However, the simple proposal distribution used in CSMC is inefficient to combine with MCMC in the framework of the particle Gibbs sampler to jointly estimate the phylogenetic tree and the evolutionary parameter. Moreover, CSMC is inapplicable to the particle Gibbs with ancestor sampling . In this talk, we will present a more efficient CSMC method, called CSMC-RMM, with a more flexible proposal. The proposed CSMC-RMM can improve the performance of the particle Gibbs sampler compared with the original CSMC. In addition, it makes the framework of particle Gibbs with ancestor sampling possible. We demonstrate the effectiveness of our new method using simulation studies and real data analysis.