As an emerging swarm intelligence optimization algorithm, the Tumbleweed Algorithm (TA) has demonstrated remarkable performance in solving complex global optimization problems. This study proposes an improved Dynamic Cauchy Tumbleweed Algorithm (DCTA), which significantly enhances the optimization performance by incorporating a Cauchy Mutation strategy and an adaptive parameter mechanism. A systematic evaluation is conducted using the CEC2017 benchmark test function set, with experiments performed in 10-, 30-, and 50-dimensional search spaces. The experimental design includes comparative experiments with the original TA algorithm, four classical swarm intelligence optimization algorithms in the literature, and two swarm intelligence algorithms that have shown excellent performance in recent years. The results demonstrate that DCTA possesses significant advantages in solving high-dimensional complex optimization problems.