Several advancements in 3D asset and molecule generation have been reported recently. DreamBeast utilizes Stable Diffusion 3 to create detailed Part-Affinity maps from various camera views, improving quality and saving computing power. Phidias generates high-quality 3D assets from text, images, and 3D references using reference-augmented diffusion, achieving results in just a few seconds. 3DTopia-XL scales high-quality 3D asset generation via primitive diffusion. FlexiTex enhances texture generation with visual guidance. StoryMaker aims towards holistic consistent characters in text-to-image generation. Additionally, a new approach enhances the quality of 3D drug-like molecule generation using diffusion models. The study introduces a novel framework, LLM-DDA, which combines large language models like GPT-4 with graph neural networks for computational drug repositioning. Another study presents a geometric self-supervised pretraining approach for 3D Graph Neural Networks to improve protein structure representation learning.