Pan Zhou

Assistant Professor@Singapore Management University·新加坡·OpenReview

研究方向

deep learning optimization · self-supervised learning · optimization algorithm · Deep learning theory analysis · including generalisation and optimisation

Pan Zhou

Distributional Vision-Language Alignment by Cauchy-Schwarz Divergence

From Spatial to Actions: Grounding Vision-Language-Action Model in Spatial Foundation Priors

Architecture-Agnostic Test-Time Adaptation via Backprop-Free Embedding Alignment

Bridging Draft Policy Misalignment: Group Tree Optimization for Speculative Decoding

Dragging with Geometry: From Pixels to Geometry-Guided Image Editing

Ego-VGA: A Compact Multimodal Assistant for Egocentric Video–Grounded Reasoning

Stacked from One: Multi-Scale Self-Injection for Context Window Extension

DreamCS: Geometry-Aware Text-to-3D Generation with Unpaired 3D Reward Supervision

Residual Pyramid Atrous Filtering Network with the Error Low-Rank Respresentation

Towards Scalable and Consistent 3D Editing

IBNorm: Information-Bottleneck Inspired Normalization for Representation Learning

2D Quantization for Ultra‑low‑bit Optimizers

VistaGUI: Towards More Robust and Intelligent GUI Automation

Towards Uniformity and Alignment for Multimodal Representation Learning

Conda: Column-Normalized Adam for Training Large Language Models Faster

Fast Inference of Visual Autoregressive Model with Adjacency-Adaptive Dynamical Draft Trees

Towards Understanding Why FixMatch Generalizes Better Than Supervised Learning

GRIFFIN: Effective Token Alignment for Faster Speculative Decoding

Probabilistic Interactive 3D Segmentation with Hierarchical Neural Processes

CaPo: Cooperative Plan Optimization for Efficient Embodied Multi-Agent Cooperation

SoPo: Text-to-Motion Generation Using Semi-Online Preference Optimization

HPS: Hard Preference Sampling for Human Preference Alignment

Two Are Better than One: Context Window Extension with Multi-Grained Self-Injection

Towards Natural Image Matting in the Wild via Real-Scenario Prior

SubZero: Random Subspace Zeroth-Order Optimization for Memory-Efficient LLM Fine-Tuning

Effective and Efficient Adversarial Detection for Vision-Language Models via A Single Vector

ActionFiller: Fill-In-The-Blank Prompting for OS Agent