Spatial
At Menlo, we’re pioneering new approaches to spatial intelligence in AI systems. Our research focuses on developing models that can understand and reason about spatial relationships, enabling more sophisticated interactions with the physical world.
Research Projects
AlphaSpace (Apr 2025)
🤖 AlphaSpace: A Step Closer to Having Clumsyless Robots
AlphaSpace is a cutting-edge framework that enhances spatial intelligence in robotics, enabling robots to perform complex tasks with precision and adaptability. By integrating advanced spatial reasoning capabilities, AlphaSpace empowers robots to navigate, plan, and execute actions in dynamic environments, achieving a 66.67% accuracy rate in manipulation tasks.
AlphaMaze (Mar 2025)
🧩 AlphaMaze: Teaching LLMs to Think Visually
AlphaMaze is an advanced decoder-only large language model designed to excel in visual reasoning tasks, particularly in solving spatial challenges like mazes. Built on a 1.5 billion parameter Qwen model, it demonstrates impressive capabilities in handling complex visual reasoning tasks through innovative techniques like GRPO.
