Manling Li
Assistant Professor
Northwestern University
Center for Robotics and Biosystems (Affiliated)
Amazon Scholar
Amazonmanling.li AT northwestern.edu
Our Research
My research vision is to fundamentally transform artificial intelligence models from passive observers into active, embodied agents. Today's foundation models, trained on static web-scale data, excel at semantic-centric reasoning, yet lack grounded interactions with the dynamic and partially observable physical world. I strive to architect this foundational bridge, like a trainable reasoning interface that connects foundation models to physical environments, addressing the core challenges in perception, policy and alignment.
Research #1: Spatial Intelligence, Embodied Intelligence, Multimodality (Language + Vision/Robotics)
Seeing the unseen: From Semantic-Centric Priors to Spatial-Geometric Reasoning.
Research #2: Reasoning, Planning and Foundation Agent Training
Exploring the world: The Self-Evolving Policy Reasoning Interface From Exploitation to Exploration via World Modeling. LLM Agents (RAGEN, Self-Play Agent),
VLM Agents (VAGEN with World Model RL), Embodied Agents (Embodied Reasoning Agent for RL training, Embodied Agent Interface for LLM agents,
EmbodiedBench for VLM agents,
ROSETTA with human feedback, etc)
Research #3: Mechanistic Interpretability, Safety, Trustworthiness
Steering the reasoning: The Mechanistic Control Interface From Black Boxes to Safe Agents.
We have a community effort on Foundation Models meet Embodied Agents
- Tutorials: LLMs, VLMs, and VLAs, we give a comprehensive overview of the research based on Markov Decision Process (MDP) framework.
- Challenges: Co-hosted with BEHAVIOR Challenge at NeurIPS 2025.
- Workshops: we will comprehensively review existing paradigms for foundations for embodied agents, and focus on their different formulations based on the fundamental mathematical framework of robot learning, Markov Decision Process (MDP), and present a structured view to investigate the robot’s decision-making process..
Research Impact
