Liwei Wang

I am an Assistant Professor in Computer Science and Engineering department at The Chinese University of Hong Kong (CUHK). Before coming to HK, I have worked for more than two years as a Senior Researcher in Tencent America at Bellevue, US.

I got my PhD from Computer Science Department, University of Illinois Urbana-Champaign (UIUC), advised by Prof. Svetlana Lazebnik. The Language and Vision (LaVi) Lab, which I founded at the Department of Computer Science and Engineering at CUHK, conducts research in Language+Vision, i.e. the intersection of language and vision.

If you want to join LaVi Lab, please send an email to lwwang@cse.cuhk.edu.hk

Email / Google Scholar / Publications / Lab website (soon) /

• We are hiring Interns, Postdocs and Phd students to work on Vision+Language and multi-modal LLMs.
• 2025/08: I will give a keynote speak at MFMSI Workshop at ACM Multimedia 2025.

My students / interns / postdocs are indicated by '*'. Click full publication list

Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors

Learning to Reason from Feedback at Test-Time

C²LEVA: Toward Comprehensive and Contamination-Free Language Model Evaluation

Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding

AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning

Fine-grained Spatiotemporal Grounding on Egocentric Videos

Towards Learning a Generalist Model for Embodied Navigation

A Mutual Supervision Framework for Referring Expression Segmentation and Generation

Making Long-Context Language Models Better Multi-Hop Reasoners

Beyond Embeddings: The Promise of Visual Table in Multi-Modal Models

Enhancing Temporal Modeling of Video LLMs via Time Gating

Learning Preference Model for LLMs via Automatic Preference Data Generation

CLEVA: Chinese Language Models EVAluation Platform

VL-PET: Vision-and-Language Parameter-Efficient Tuning via Granularity Control

Multi-View Transformer for 3D Visual Grounding

Eliciting Knowledge from Large Pre-Trained Models for Unsupervised Knowledge-Grounded Conversation

FlowEval: A Consensus-Based Dialogue Evaluation Framework Using Segment Act Flows

Probing Structured Pruning on Multilingual Pre-trained Models: Settings, Algorithms, and Efficiency

SAT: 2D Semantics Assisted Training for 3D Visual Grounding

Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation

Robust Dialogue Utterance Rewriting as Sequence Tagging

Comprehensive Image Captioning via Scene Graph Decomposition

Improving One-stage Visual Grounding by Recursive Sub-query Construction

MART: Memory-Augmented Recurrent Transformer for Coherent Video Paragraph Captioning

A Fast and Accurate One-Stage Approach to Visual Grounding​

Fast, Diverse and Accurate Image Captioning Guided By Part-of-Speech

Learning Two-Branch Neural Networks for Image-Text Matching Tasks

Learning structural motif representations for efficient protein structure search

Diverse and Accurate Image Description Using a Variational Auto-Encoder with an Additive Gaussian Encoding Space

Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models

Learning Deep Structure-Preserving Image-Text Embeddings

(Templates adapted from Arun's webpage.)