Aiming at two-stage reverse supply chain which consists of manufacturers and third-party recyclers, an evolutionary game model of knowledge-sharing behavior among enterprises in supply chain was established by using evolutionary game method. The equilibriums, optimal strategy choice and influencing factors of knowledge-sharing among enterprises in reverse supply chain were analyzed. The results show that the evolution system of knowledge-sharing behavior selection among enterprises in reverse supply chain can converge not only to the rational state of knowledge sharing, but also to the bad "lock in" state of knowledge non-sharing. This is mainly related to the amount of knowledge gaps, incentives, knowledge absorption and transformation capabilities, sharing risk and sharing costs among enterprises in reverse supply chain. According to the analysis conclusions, several suggestions are proposed from the aspects of enterprise scale, enterprise learning ability, internal and external incentive mechanism and trust mechanism in reverse supply chain.