<p>Multi-task optimization algorithms are widely recognized for their effectiveness in addressing complex problems. However, they frequently encounter issues such as negative knowledge transfer, reduced convergence efficiency, and the risk of being trapped in local optima throughout the optimization process. To overcome these difficulties, this paper proposes an improved adaptive multi-tasking differential evolutionary algorithm with opposition-based learning (OBLMTDE). The algorithm achieves knowledge balance between tasks by adaptively adjusting the random mating probability, thus accelerating the convergence speed. At the same time, a mutation strategy based on elite selection is introduced, which promotes positive knowledge transfer by selecting excellent individuals engaged in mutation in other tasks. In addition, the algorithm adopts a novel parameter tuning strategy that enhances the global exploration capability and effectively avoids the local optimum trap. To validate the performance of the OBLMTDE algorithm, we conducted experimental comparisons on nine benchmark problems of single-objective multi-task optimization. Our experimental results indicate that OBLMTDE outperforms existing multi-tasking optimization algorithms in most of the tested functions.</p>