Decoding the Multimodal Maze: A Systematic Review on the Adoption of Explainability in Multimodal Attention-based Models

This systematic review analyzes the adoption of explainability in multimodal attention-based models, focusing on literature from January 2020 to early 2024. It highlights the dominance of vision-language and language-only models, critiquing the limitations of current attention-based explanation methods in capturing inter-modal interactions and the lack of systematic evaluation methodologies. The findings underscore the necessity for standardized practices in multimodal explainability research, aiming to enhance interpretability and accountability in AI systems.