Currently, I am interested in several topics, including but not limited to:
- Native Multimodal Foundation Model: Rather than retrofitting language models with vision or audio adapters, native multimodal foundation models are designed from the ground up to jointly perceive and reason across text, images, video, and audio within a single architecture. This paradigm treats every modality as a first-class citizen during pre-training, enabling richer cross-modal representations, more coherent multimodal reasoning, and emergent capabilities that modular pipelines struggle to achieve. Core challenges include designing unified tokenization schemes, balancing modality-specific and shared representations, crafting scalable pre-training objectives that capture inter-modal dependencies, and establishing principled scaling laws that govern how performance evolves as data, compute, and the number of modalities grow.
- Efficient & Scalable AI: As foundation models continue to scale in both parameter count and training data, many conventional techniques face diminishing returns or become prohibitively expensive. Advancing efficient and scalable methodologies—spanning adaptive optimization strategies, speculative and parallel decoding, low-precision training, memory-efficient attention mechanisms, and domain-specific acceleration techniques for generative models—is essential to sustaining progress. Crucially, these approaches must not only reduce computational cost at today's scale but also preserve their effectiveness as model size, sequence length, and deployment complexity continue to grow.
- World Model: World models seek to learn an internal, predictive representation of the environment that can simulate future states, reason about physical dynamics, and support planning without exhaustive trial-and-error interaction. By capturing spatial, temporal, and causal structure from large-scale video and sensor data, world models hold the potential to serve as a general-purpose "mental simulator" for embodied agents, autonomous driving, robotics, and scientific discovery. Central challenges include learning physically grounded dynamics from passive observation, generalizing across diverse environments and embodiments, achieving long-horizon temporal consistency, and efficiently integrating world-model predictions into downstream decision-making and control.
- Agent: Agents extend foundation models from passive prediction systems into active problem solvers that can perceive context, plan over long horizons, use tools, interact with environments, and learn from feedback. By integrating strong base models with memory, reasoning, action, and verification mechanisms, agentic systems can push these models toward a higher ceiling of real-world capability while revealing where their intelligence succeeds or breaks down. Studying agents therefore offers both a practical path toward more capable applications and a rigorous way to explore the boundaries of intelligence, including autonomy, adaptability, reliability, and generalization in open-ended environments.
The following illustrates how two complementary research pillars — Efficient & Scalable AI and Generative AI — jointly converge toward Multimodal Logical Intelligence.
"Efficiency isn't about doing less — it's about understanding more."
"What I cannot create, I do not understand."
Developing optimizer-architecture co-design principles and scalable training strategies that remain effective across diverse model scales.
Developing agentic systems that autonomously plan, generate, evaluate, and iteratively refine multimodal content through reasoning, tool use, and feedback-driven interaction.
Designing scalable AI infrastructure that unifies training and inference to bridge the training–serving gap and enable efficient, reliable, and low-latency deployment.
High-fidelity generation across images, video, 3D/4D, and beyond — advancing quality, consistency, and compositional control.
Designing model architectures with inherent scalability — exploring structural inductive biases, modular composition, and architecture-optimizer synergies that generalize as capacity grows.
Learning predictive, physically grounded representations of environments to enable simulation, planning, and embodied reasoning.
A next frontier beyond multimodal perception and generation: intelligence that connects evidence across text, images, video, audio, and 3D; performs compositional, causal, and counterfactual reasoning; and transforms multimodal understanding into coherent, verifiable decisions and actions in the physical world.