Large Language Models and Embodied Intelligence

Aims:

The Large Language Models and Embodied Intelligence Section is dedicated to the frontier of AI research: the creation of large-scale, foundational models and their integration into physical or simulated worlds. This Section has two interconnected pillars. First, it focuses on the development, scaling, and societal impact of large models (e.g., for language, vision, and multimodal data). Second, it explores embodied intelligence—how these and other AI models can be deployed in robots, avatars, and digital twins to enable intelligent interaction with a dynamic human-centered environment.

Scope:

This Section captures the most recent and impactful trends in cognitive computing. The scope includes the following:

• Large Language Models (LLMs) and Foundation Models: Architecture scaling, pre-training, fine-tuning, alignment, and capabilities.
• Multimodal Large Models: Systems that process and integrate vision, language, audio, and other sensory data.
• Natural Language Processing (NLP): Advanced applications of large models in translation, summarization, and dialogue.
• Embodied AI and Cognitive Robots: Algorithms for robot perception, navigation, and manipulation that require physical grounding.
• Digital Twin Technologies: Creating and using dynamic virtual models of physical systems for simulation, analysis, and control.
• 6G Semantic Communication: Task-oriented and knowledge-aware communication networks for agents and large models.
• AI Safety and Alignment in Large Models: Research on robustness, controllability, and the societal impact of powerful generative models.