Abstract: To investigate the bond performance between high strength steel rebar and ultra-high-performance concrete (UHPC), 17 sets of specimens were designed for center pull-out tests. The failure modes of high-strength steel bar embedded in UHPC and the effects of rebar diameter, of relative protective layer thickness, of steel fiber volume content, of rebar strength and, of anchorage length on the bond performance were analyzed through the test. The experimental results show that specimens with smaller-diameter rebars (16~20 mm) exhibit higher average bond strength and peak slip in comparison to those with larger-diameter rebars (22~25 mm). The average bond strength of HRB500 rebars in UHPC is 26.2% higher than that of HRB400 rebars. As the steel fiber volume content is enhanced from 1% to 3%, the average bond strength first increases and then decreases, reaching its maximum at a volume content of 2%. The average bond strength increases with the UHPC cover thickness. However, when the cover thickness exceeds 4 times the rebar diameter, the growth of the average bond strength becomes negligible. As the bonding length increases from 2 times the rebar diameter to 4 times or more, the failure mode of rebar gradually changes from pulled out with no sign of yielding to tensile fracture, and the average bond strength decreases. To predict the bond strength between rebar and UHPC, a database of 325 sets on bond strength was established, and two machine learning models were developed: the BP neural network optimized by genetic algorithm (GA-BP) model and the random forest (RF) model. Both machine learning models demonstrate a good consistency with test results, among which the GA-BP exhibits superior fitting accuracy and an improved prediction stability.