The extraction of induced polarization (IP) information from transient electromagnetic (TEM) signals holds significant practical importance for the development of deep mineral, oil and gas resources. Linear inversion technology is the preferred method for extracting IP information, but it is associated with three primary drawbacks: dependence on the initial conditions, susceptibility to falling into a local optimum, and a significant lack of uniqueness. To solve the above problems, an improved shuffle frog leaping algorithm (ISFLA) based on the tent chaotic distribution and an adaptive mobile factor is presented in this paper, and the algorithm is employed to extract IP information. First, a tent chaotic operator is adopted to ameliorate the initial population distribution to improve the global search capability. Then, an adaptive mobile factor is designed to replace the random operator for balancing the local and global search, which increases the solution accuracy and ensures stable convergence in the later period. Finally, TEM inversion for a 1-D layered geoelectric model with IP information is performed by the proposed ISFLA approach. The inversion results show that the ISFLA method can more effectively reconstruct the geoelectric structure as well as extract the IP information and achieve stronger robustness. Compared with other heuristic algorithm, the proposed algorithm achieves a superior global search ability and inversion accuracy, making it suitable for IP information extraction.