A Benchmark Study of Hybrid CNN-Transformer Architectures in Vision-Language Tasks. Emerging Science Research 2025, 3 (01), 36-49.