Brain X is a robot action-model framework that combines robot foundation models,

reinforcement learning, and imitation learning to enable practical task execution across diverse environments.

Brain X consists of a conversational action model, hierarchical/compositional

reinforcement learning models, and behavior imitation models.

BrainX is composed of System 1 and System 2, designed to switch dynamically

depending on task complexity. In System 2, a multimodal robot agent performs

perception and reasoning, enabling real-world object manipulation and human interaction.

BRAIN X orchestrates sub-action agents with Reinforcement Learning (Hierarchical RL).

By decomposing the reinforcement learning policy into sub-policies,

we can build a structure that is more robust to generalization tasks and improves task success rates.

Conversational Action Model

Our conversational Action Model, Voice X™, enables robots to engage in verbal

communication seamlessly integrated with physical responses, leveraging multimodal embeddings

that generate both natural dialogue and coordinated robot actions.

Human Motion Mimicel

Our Human Motion Imitation technology support Humanoid motion retargeting with reference video motion.

Here is an example of extracting a person's dancing from a video and applying them to a robot.

Imitation / Reinforcement Learning

At XYZ, we focus on commercializing Imitation Learning and Reinforcement Learning so they can be used in everyday life.

The model applied to Baris Brew is responsible for handling situations where a cup is accidentally knocked over.

Imitation / Reinforcement Learning

At XYZ, we focus on commercializing Imitation Learning and Reinforcement Learning so they can be used in everyday life.

The model applied to Baris Brew is responsible for handling situations where a cup is accidentally knocked over.

Physical Data

Collection Device

- Our Glove X™ achieve and distill millions of human movements into its onboard AI, enabling helpful actions for dailylife.

- Our Glass X™ record millions of human manipulations into its onboard AI, building action models for dailylife.

Brain X is a robot action-model framework

that combines robot foundation models,

reinforcement learning, and imitation learning to enable practical task execution across diverse environments.

Brain X consists of a conversational action model, hierarchical/compositional reinforcement learning models,

and behavior imitation models.

BrainX is composed of System 1 and System 2,

designed to switch dynamically

depending on task complexity.

In System 2, a multimodal robot agent performs

perception and reasoning, enabling real-world object manipulation and human interaction.

BRAIN X orchestrates sub-action agents

with Reinforcement Learning (Hierarchical RL).

By decomposing the reinforcement learning policy

into sub-policies, we can build a structure that is more robust to generalization tasks and improves task success rates.

Conversational Action Modelc

Our conversational Action Model, Voice X™, enables robots to engage in verbal communication seamlessly integrated with physical responses, leveraging multimodal embeddings that generate both natural dialogue and coordinated robot actions.

Human Motion Mimic

Our Human Motion Imitation technology support Humanoid motion retargeting with reference video motion.

Here is an example of extracting a person's dancing from a video and applying them to a robot.

Imitation / Reinforcement Learning

At XYZ, we focus on commercializing Imitation Learning and Reinforcement Learning so they can be used in everyday life.

The model applied to Baris Brew is responsible for handling situations where a cup is accidentally knocked over.

Digital Twin Platform

for Operating Remote

Robots in Real-time

TwinX™ is a virtual environment platform synchronized with the real world in real time, enabling human

operators to remotely control robots located at a distance via a leader interface or a web interface.

These series of tasks form human behavior data for the LWM (Large World Model).

Physical Data

Collection Device

- Our Glove X™ achieve and distill millions of human movements into its onboard AI, enabling helpful actions

for dailylife.

- Our Glass X™ record millions of human manipulations into its onboard AI, building action models for dailylife.