Researchers proposed a plant growth and shade dataset and a generative shade simulation model using diffusion models. The simulation model can generate realistic shades for unseen plants and supports downstream robotic tasks like perception, lighting control, and view planning.

This research paper investigates the theoretical limits of reliability and scaling in large language models (LLMs). It challenges the assumption that perfect reliability is achievable with sufficient scale and proposes a first-principles scaling law that explains the relationship between model performance, training data, and model capacity. The findings have implications for the development and deployment of LLMs, including the potential benefits of retrieval-augmentation and the tradeoffs between model size and training data.

Researchers introduced SOLAR, a multimodal generative model for tomato disease analysis. SOLAR integrates visual and textual information to provide a comprehensive understanding of tomato leaf disease. The model outperforms state-of-the-art models in accuracy, robustness, and multimodal reasoning.

A new AI model, combining denoising diffusion and adversarial training, generates 3D geological microstructures from 2D images with high accuracy and minimal artifacts. This innovation could improve materials science and geological waste disposal, but its direct impact on AI agents is limited to potential applications in generative modeling and simulation.

Researchers introduce GenStream, a semantic streaming framework that reduces video bandwidth by transmitting structured metadata instead of dense video frames. This allows for extreme compression (99.9% reduction) and enables new directions in volumetric avatar synthesis, 3D actor fusion, and personalized viewing experiences.

Researchers developed a multimodal generative framework using Stable Diffusion XL and LLaMA to create controllable and culturally faithful Ulos motif designs. The framework incorporates four conditioning mechanisms: text, image, representation, and semantic map. An ablation study showed that combining all four mechanisms resulted in the weakest FID score, indicating conflicting optimization signals. The framework was evaluated by weavers and the public, with statistically significant positive acceptance.