Bare-bones particle swarm optimization (BPSO) converges quickly and is parameter-free. But BPSO is easily suffer from the premature convergence. This paper presents a distribution-guided BPSO (DBPSO) in which an adaptive jump operation is introduced to help the particle get out of the local optimal and each dimension of the particle is assigned a jump probability according to the evolutionary state of the swarm. The new jump operation can improve the ability of escaping the local optima. The proposed DBPSO has been experimentally validated on 17 benchmark functions. Compared with several BPSO variants, experimental results and statistic analysis show DBPSO is competitive in solving most functions.