The field of text-to-3D generation and scene understanding is rapidly evolving, with a focus on improving the logical coherence, spatial interactions, and adaptability of generated scenes. Recent developments have highlighted the importance of incorporating causal reasoning, vision-language models, and structured information to enhance the quality and accuracy of generated 3D scenes and images. Notably, the integration of large language models and vision-language models has shown promising results in addressing the challenges of semantic fidelity, geometric coherence, and spatial correctness. Furthermore, the use of tuple-based structured information and knowledge distillation techniques has demonstrated significant improvements in spatial accuracy and action depiction in text-to-image generation. Overall, these advancements are pushing the boundaries of what is possible in text-to-3D generation and scene understanding, enabling more realistic and contextually accurate outputs. Noteworthy papers include: CausalStruct, which proposes a novel framework for controllable 3D scene generation using causal reasoning and large language models. VLM3D, which integrates vision-language models into the SDS pipeline to improve semantic fidelity and geometric coherence. AcT2I, which introduces a benchmark for evaluating action depiction in text-to-image models and proposes a knowledge distillation technique to address the limitation of current T2I methods.