Research Overview

Our research has broadly focused on the following fields:

1. Spatial AI & Computing

• [CVPR'25] GaussianSpa: An “Optimizing-Sparsifying” Simplification Framework for Compact and High-Quality 3D Gaussian Splatting
  
• [NeurIPS'23] GraphMP: Graph Neural Network-based Motion Planning with Efficient Graph Search
  
• [IROS'22] Robot Motion Planning as Video Prediction: A Spatio-Temporal Neural Network-based Motion Planner
  

2. AI for Scientific Computing & Data Management

• [Preprint'25] FLARE: A Dataflow-Aware and Scalable Hardware Architecture for Neural-Hybrid Scientific Lossy Compression
  
• [HPDC'25] Advancing Scientific Data Compression via Cross-Field Prediction
  
• [ICS'25] NeurLZ: An Online Neural Learning-based Method to Enhance Scientific Lossy Compression
  
• [DRBSD@SC'24] Enhancing Lossy Compression Through Cross-Field Information for Scientific Applications
  
• [FlexScience@HPDC'24] GWLZ: A Group-wise Learning-based Lossy Compression Framework for Scientific Data
  

3. Multimodal Media Intelligence

• [CVPR'25 Highlight] AdaCM2: On Understanding Extremely Long-Term Video with Adaptive Cross-Modality Memory Reduction
  

4. AI/ML Systems & Hardware

• [ASPLOS'26] FlashMem: Supporting Modern DNN Workloads on Mobile with GPU Memory Hierarchy Optimizations
  
• [NeurIPS'24] Real-time Core-Periphery Guided ViT with Smart Data Layout Selection on Mobile Devices
  
• [ASPLOS'24] SmartMem: Layout Transformation Elimination and Adaptation for Efficient DNN Execution on Mobile
  
• [ISCA'23] ETTE: Efficient Tensor-Train-based Computing Engine for Deep Neural Networks
  
• [PPoPP'23] TDC: Towards Extremely Efficient CNNs on GPUs via Hardware-Aware Tucker Decomposition
  
• [TC'22] Algorithm and Hardware Co-Design of Energy-Efficient LSTM Networks for Video Recognition with Hierarchical Tucker Tensor Decomposition
  
• [ICCAD'19] (Invited Paper) High-performance Hardware Architecture for Tensor Singular Value Decomposition
  

5. AI/ML Model Optimization/Compression/Adaptation

• [ICLR'26] LeSTD: LLM Compression via Learning-based Sparse Tensor Decomposition
  
• [Preprint'25] AdaRing: Towards Ultra-Light Vision-Language Adaptation via Cross-Layer Tensor Ring Decomposition
  
• [EMNLP'24 Findings] MoE-I2: Compressing Mixture of Experts Models through Inter-Expert Pruning and Intra-Expert Low-Rank Decomposition
  
• [ICML'23] COMCAT: Towards Efficient Compression and Customization of Attention-Based Vision Models
  
• [AAAI'23] CSTAR: Towards Compact and Structured Deep Neural Networks with Adversarial Robustness
  
• [AAAI'23] GOHSP: A Unified Framework of Graph and Optimization-based Heterogeneous Structured Pruning for Vision Transformer
  
• [CVPR'22] HODEC: Towards Efficient High-Order DEcomposed Convolutional Neural Networks
  
• [AAAI'22] BATUDE: Budget-Aware Neural Network Compression Based on Tucker Decomposition
  
• [NeurIPS'21] CHIP: CHannel Independence based Pruning for Compact Neural Networks
  
• [CVPR'21] Towards Efficient Tensor Decomposition-Based DNN Model Compression with Optimization Framework
  
• [CVPR'21] Towards Extremely Compact RNNs for Video Recognition with Fully Decomposed Hierarchical Tucker Structure