The field of AI-driven systems is witnessing a significant shift towards integrating causality and structural insights to provide more interpretable and transparent results. Researchers are exploring novel approaches to combine causal analysis with large language models (LLMs) to generate human-readable explanations for complex phenomena. Furthermore, there is a growing interest in designing inductive biases for document recognition systems to capture the intrinsic structure of documents, enabling more accurate and efficient transcription. Additionally, the development of substructure reasoning in transformers is opening up new avenues for graph reasoning and extraction tasks. Noteworthy papers include:

• InsightBuild, which proposes a two-stage framework for causal reasoning in smart building systems.
• NGTM, which introduces a novel generative framework for interpretable graph generation.