Topics

Bio-inspired adaptive neurons that dynamically adjust weights during inference, bringing biological flexibility to rigid artificial neural networks for smarter, more adaptable AI.

Discover how Text-to-LoRA (T2L) enables instant, zero-shot adaptation of LLMs—compressing hundreds of LoRA adapters, generating task-specific weights from language, and personalizing AI behavior effortlessly.

NEAT evolves neural networks like living organisms, starting simple, protecting innovation, and growing smarter architectures that learn dramatically faster through artificial evolution.

Neural Cellular Automata scaled to HD: Coarse NCA handles self-organizing structure on a small grid, while lightweight LPPN renders high-resolution details in real-time for images, textures, and 3D meshes.

Explore a revolutionary neural network architecture integrating Multi-Query, Multi-Head, and Grouped-Query Attention, enhancing efficiency and adaptability in multimodal tasks while reducing computational costs and memory usage.

A 150-million-parameter system scores nearly 30 percent on a public intelligence benchmark while costing a fraction of larger models that rely on written reasoning steps.

Researchers develop a new type of neural network that learns continually with much lower power consumption than today's large AI models.

DeepSeek's R1 has shocked the tech world and disrupted the stock market with the astonishingly low cost of development they claimed.