Research
I am interested in end-to-end hardware-software co-design. My publications span AR/VR, AI, and imaging systems.
I am also an AI maximalist interested in Autoresearch/AI4S, aiming to accelerate scientific discovery by orders of magnitude and unlock research that was previously impossible.
AR/VR Systems
Building efficient AR/VR systems through joint optimization across rendering algorithms, displays, hardware, and perception. This includes power-aware tone mapping, real-time Gaussian splatting, foveated rendering, and streaming architectures for immersive content.
Related publications
- arXiv 2026GLoD-LVSM: Learning Geometry-Grounded Continuous-LoD Scene Representations for Large View Synthesis Models
- SIGGRAPH Asia 2026ControlGS: Conditioning Neural Gaussians for Downstream-Processing-Aware XR Rendering
- arXiv 2026MetaSapiens v2: Advancing Real-Time Foveated Neural Rendering via Foveation-Aware Pruning and Stereo Warping
- TVCG 2026LowPowAR: Power-Constrained Tone Mapping for Augmented Reality
- CVPR 2026SeeLe: A Unified Acceleration Framework for Real-Time Gaussian Splatting on Mobile Devices Highlight
- SIGGRAPH Asia 2025PowerGS: Display-Rendering Power Co-Optimization for Neural Rendering in Power-Constrained XR Systems
- ASPLOS 2025MetaSapiens: Real-Time Neural Rendering with Efficiency-Aware Pruning and Accelerated Foveated Rendering Best Paper Award
- ASPLOS 2025StreamGrid: Streaming Point Cloud Analytics via Compulsory Splitting and Deterministic Termination
- ISCA 2025Lumina: Real-Time Mobile Neural Rendering by Exploiting Computational Redundancy
- HotMobile 2025Advancing Immersive Content Delivery with Dynamic 3D Gaussian Splatting
- ACM TACO 2024Potamoi: Accelerating Neural Rendering via a Unified Streaming Architecture
- arXiv 2024RTGS: Enabling Real-Time Gaussian Splatting on Mobile Devices Using Efficiency-Guided Pruning and Foveated Rendering
- Project 2024Exploiting Human Color Discrimination for Memory and Energy-Efficient Image Encoding in Virtual Reality: An FPGA Demo
AI Systems
Developing artificial intelligence across generative and multimodal learning, embodied perception, and robot learning. This work spans controllable visual generation, cross-modal alignment, open-world perception, and autonomous systems.
Related publications
- ECCV 2026ControlHair: Synergizing Physics Simulator and Video Diffusion for Controllable Dynamic Hair Rendering
- NeurIPS 2026GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction
- NeurIPS 2026Multidimensional Observer Model and Perceptual Dimensions of Human Image Quality Assessment
- ACM MM 2026UniMod: Enhancing Multi-Modal Medical Diagnosis through Cross-Modality and Within-Modality Alignment
- EMNLP 2026The Menu Is an Execution Prior: State-Path Tool Menus for Online Agents
- arXiv 2026MemRouter: Memory-as-Embedding Routing for Long-Term Conversational Agents
- arXiv 2026SafeDream: Safety World Model for Proactive Early Jailbreak Detection
- IROS 2024OW3Det: Toward Open-World 3D Object Detection for Autonomous Driving
- ROMAN 2023Learning Clear Class Separation for Open-set 3D Detector in Autonomous Vehicle via Selective Forgetting
- PKU ASN 2023A Review of Robot Learning
- ICDL 2021Approaching Sound Object with Sensorimotor Coordination when Sensors Partially Damaged
- ICDL 2021Acquiring Robot Navigation Skill with Knowledge Learned from Demonstration
Imaging Systems
Designing sensors and optics for imaging systems, along with co-optimized downstream vision models. This involves in-sensor compression inspired by efficient coding theory and privacy-preserving optical feature separation.
Related publications
- DAC 2026HoloCode: Hybrid Optical-Electronic Edge Encoding for Privacy-Preserving Cloud Training
- DAC 2026Invited Paper: A Quantitative Approach to Exploring Novel Image Sensor Architecture Towards Autonomous Edge Machine Vision
- DAC 2025SnapPix: Efficient-Coding-Inspired In-Sensor Compression for Edge Vision
- WACV 2025PrivateEye: In-Sensor Privacy Preservation Through Optical Feature Separation