<p>This work introduces a novel approach for autonomous rule modeling in evolutionary neuro-fuzzy systems. The proposed model, called DGP-NF, combines multi-gene genetic programming with dynamic population and diversity control strategies, integrated with an incremental learning algorithm that adjusts network parameters. DGP-NF simultaneously evolves the fuzzy rule base structure and network parameters, aiming to improve predictive accuracy and convergence behavior. The model uses population diversity measures based on phenotypic, rather than genotypic, characteristics, since phenotypes directly influence the system output and, consequently, its predictive performance. Furthermore, a pruning strategy for multi-gene individuals, guided by phenotypic diversity, is introduced to reduce redundancy in the evolved rule base. Experimental results on 21 non-linear regression datasets show that DGP-NF outperforms alternative models on 16 datasets and achieves competitive performance on the remaining ones, indicating generalization capability. The use of dynamic population and diversity control mechanisms proves effective in mitigating premature convergence and promoting broader exploration of the search space. Additionally, the proposed diversity-based pruning strategy demonstrates potential as a simple and effective approach for refining multi-gene individuals.</p>