邓画予Huayu Deng
PhD student, Shanghai Jiao Tong University · Research Intern, RedNote
Shanghai, China
I am a PhD student in the Wu Wenjun Honors Class (吴文俊人工智能荣誉博士班) in Artificial Intelligence at Shanghai Jiao Tong University, advised by Yunbo Wang and Xiaokang Yang. Before that, I completed my B.Eng. in Computer Science (IEEE Honors Class) at Shanghai Jiao Tong University. I am currently a research intern at RedNote (小红书), working on large language model post-training.
My research spans LLM Post-training & Embodied Interaction and AI for Physics:
- LLM Post-training & Embodied Interaction
- On-policy distillation and reinforcement learning
- Test-time training and long-context reasoning for LLMs
- Transferring interaction knowledge from video to physical control
- AI for Physics
- Active sensing and world models for continuum field reconstruction
- Learning hidden physical dynamics from visual observations
- Adaptive, multi-scale simulation of physical systems
Full publications are available on Google Scholar or Publications (* denotes equal contribution).
Works are organized by topic: AI4Physics, LLM & Embodied AI.
Works are organized by topic: AI4Physics, LLM & Embodied AI.
-
Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation
-
D³-MOPD: Adaptive Dynamic Domain ScheDuling for Efficient Multi-Teacher Distillation
-
TTT-Sparse: Test-Time Training of Sparse Attention for Long-Context Understanding
-
LASER: Learning Active Sensing for Continuum Field Reconstruction
-
Learning Transferable Interaction Primitives from Game Videos for Humanoids
-
EvoMesh: Adaptive Physical Simulation with Hierarchical Graph Evolutions
-
Latent Intuitive Physics: Learning to Transfer Hidden Physics from A 3D Video
-
Neurofluid: Fluid dynamics grounding with particle-driven neural radiance fields