Publications (47)
High-Fidelity Two-Step Image Generation via Teacher-Aligned End-to-End Distillation
Dongyang Liu, Ruoyi Du, David Liu +7
Few-step diffusion distillation has become increasingly mature for 4-8-step generation, yet pushing further to 2 steps remains challenging. In this work, we introduce Z-Image Turbo…
BiaScope: Visual Unfairness Diagnosis for Graph Embeddings
Agapi Rissaki, Bruno Scarone, David Liu +4
The issue of bias (i.e., systematic unfairness) in machine learning models has recently attracted the attention of both researchers and practitioners. For the graph mining communit…
Deep-Learning-based Fast and Accurate 3D CT Deformable Image Registration in Lung Cancer
Yuzhen Ding, Hongying Feng, Yunze Yang +9
Purpose: In some proton therapy facilities, patient alignment relies on two 2D orthogonal kV images, taken at fixed, oblique angles, as no 3D on-the-bed imaging is available. The v…
Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer
Image Team, Huanqia Cai, Sihan Cao +21
The landscape of high-performance image generation models is currently dominated by proprietary systems, such as Nano Banana Pro and Seedream 4.0. Leading open-source alternatives,…
Core-Periphery Principle Guided Redesign of Self-Attention in Transformers
Xiaowei Yu, Lu Zhang, Haixing Dai +6
Designing more efficient, reliable, and explainable neural network architectures is critical to studies that are based on artificial intelligence (AI) techniques. Previous studies,…
MMViT: Multiscale Multiview Vision Transformers
Yuchen Liu, Natasha Ong, Kaiyan Peng +8
We present Multiscale Multiview Vision Transformers (MMViT), which introduces multiscale feature maps and multiview encodings to transformer models. Our model encodes different vie…
Representation-Level Adversarial Regularization for Clinically Aligned Multitask Thyroid Ultrasound Assessment
Dina Salama, Mohamed Mahmoud, Nourhan Bayasi +2
Thyroid ultrasound is the first-line exam for assessing thyroid nodules and determining whether biopsy is warranted. In routine reporting, radiologists produce two coupled outputs:…
Mono2D: A Trainable Monogenic Layer for Robust Knee Cartilage Segmentation on Out-of-Distribution 2D Ultrasound Data
Alvin Kimbowa, Arjun Parmar, Maziar Badii +3
Automated knee cartilage segmentation using point-of-care ultrasound devices and deep-learning networks has the potential to enhance the management of knee osteoarthritis. However,…
MonoUNet: A Robust Tiny Neural Network for Automated Knee Cartilage Segmentation on Point-of-Care Ultrasound Devices
Alvin Kimbowa, Arjun Parmar, Ibrahim Mujtaba +5
Objective: To develop a robust and compact deep learning model for automated knee cartilage segmentation on point-of-care ultrasound (POCUS) devices. Methods: We propose MonoUNet,…
Bypassing Skip-Gram Negative Sampling: Dimension Regularization as a More Efficient Alternative for Graph Embeddings
David Liu, Arjun Seshadri, Tina Eliassi-Rad +1
A wide range of graph embedding objectives decompose into two components: one that enforces similarity, attracting the embeddings of nodes that are perceived as similar, and anothe…
Human Curation and Convnets: Powering Item-to-Item Recommendations on Pinterest
Dmitry Kislyuk, Yuchen Liu, David Liu +2
This paper presents Pinterest Related Pins, an item-to-item recommendation system that combines collaborative filtering with content-based ranking. We demonstrate that signals deri…
Hierarchical Semantic Tree Concept Whitening for Interpretable Image Classification
Haixing Dai, Lu Zhang, Lin Zhao +9
With the popularity of deep neural networks (DNNs), model interpretability is becoming a critical concern. Many approaches have been developed to tackle the problem through post-ho…
Ab-initio multimode linewidth theory for arbitrary inhomogeneous laser cavities
Adi Pick, Alex Cerjan, David Liu +4
We present a multimode laser-linewidth theory for arbitrary cavity structures and geometries that contains nearly all previously known effects and also finds new nonlinear and mult…
DT/MARS-CycleGAN: Improved Object Detection for MARS Phenotyping Robot
David Liu, Zhengkun Li, Zihao Wu +1
Robotic crop phenotyping has emerged as a key technology to assess crops' morphological and physiological traits at scale. These phenotypical measurements are essential for develop…
normflows: A PyTorch Package for Normalizing Flows
Vincent Stimper, David Liu, Andrew Campbell +4
Normalizing flows model probability distributions through an expressive tractable density. They transform a simple base distribution, such as a Gaussian, through a sequence of inve…
EngThrive: Make It Fast and Easy to Do Great Work
Brian Houck, Tim Bozarth, David Liu +1
Frameworks such as SPACE, DevEx, and DORA established that developer productivity is inherently multidimensional, but left practitioners with a practical question: what should we m…
Eye-gaze-guided Vision Transformer for Rectifying Shortcut Learning
Chong Ma, Lin Zhao, Yuzhong Chen +15
Learning harmful shortcuts such as spurious correlations and biases prevents deep neural networks from learning the meaningful and useful representations, thus jeopardizing the gen…
Automatic Vertebra Labeling in Large-Scale 3D CT using Deep Image-to-Image Network with Message Passing and Sparsity Regularization
Dong Yang, Tao Xiong, Daguang Xu +10
Automatic localization and labeling of vertebra in 3D medical images plays an important role in many clinical tasks, including pathological diagnosis, surgical planning and postope…
Global PIQA: Evaluating Commonsense Reasoning Across 100+ Languages and Cultures
Tyler A. Chang, Catherine Arnett, Abdelrahman Sadallah +377
To date, there exist almost no culturally-specific evaluation benchmarks for large language models (LLMs) that cover a large number of languages and cultures. In this paper, we pre…
Pebbling Arguments for Tree Evaluation
David Liu
The Tree Evaluation Problem was introduced by Cook et al. in 2010 as a candidate for separating P from L and NL. The most general space lower bounds known for the Tree Evaluation P…
Identifying and Upweighting Power-Niche Users to Mitigate Popularity Bias in Recommendations
David Liu, Erik Weis, Moritz Laber +2
Recommender systems have been shown to exhibit popularity bias by over-recommending popular items and under-recommending relevant niche items. We seek to understand niche users in…
Evolving Antennas for Ultra-High Energy Neutrino Detection
Julie Rolla, Amy Connolly, Kai Staats +16
Evolutionary algorithms borrow from biology the concepts of mutation and selection in order to evolve optimized solutions to known problems. The GENETIS collaboration is developing…
Universal Witnesses for State Complexity of Basic Operations Combined with Reversal
Janusz Brzozowski, David Liu
We study the state complexity of boolean operations, concatenation and star with one or two of the argument languages reversed. We derive tight upper bounds for the symmetric diffe…
RAWLSNET: Altering Bayesian Networks to Encode Rawlsian Fair Equality of Opportunity
David Liu, Zohair Shafi, William Fleisher +2
We present RAWLSNET, a system for altering Bayesian Network (BN) models to satisfy the Rawlsian principle of fair equality of opportunity (FEO). RAWLSNET's BN models generate aspir…
Transformation Inverse Design
David Liu, Lucas H. Gabrielli, Michal Lipson +1
We present a new technique for the design of transformation-optics devices based on large-scale optimization to achieve the optimal effective isotropic dielectric materials within…
When Collaborative Filtering is not Collaborative: Unfairness of PCA for Recommendations
David Liu, Jackie Baek, Tina Eliassi-Rad
We study the fairness of dimensionality reduction methods for recommendations. We focus on the fundamental method of principal component analysis (PCA), which identifies latent com…
ReFuzz: Reusing Tests for Processor Fuzzing with Contextual Bandits
Chen Chen, Zaiyan Xu, Mohamadreza Rostami +4
Processor designs rely on iterative modifications and reuse well-established designs. However, this reuse of prior designs also leads to similar vulnerabilities across multiple pro…
Forecasting Faculty Placement from Patterns in Co-authorship Networks
Samantha Dies, David Liu, Tina Eliassi-Rad
Faculty hiring shapes the flow of ideas, resources, and opportunities in academia, influencing not only individual career trajectories but also broader patterns of institutional pr…
ViThinker: Active Vision-Language Reasoning via Dynamic Perceptual Querying
Weihang You, Qingchan Zhu, David Liu +3
Chain-of-Thought (CoT) reasoning excels in language models but struggles in vision-language models due to premature visual-to-text conversion that discards continuous information s…
Interaction-induced mode switching in steady-state microlasers
Li Ge, David Liu, Alexander Cerjan +5
We demonstrate that due to strong modal interactions through cross-gain saturation, the onset of a new lasing mode can switch off an existing mode via a negative power slope. In th…
Group fairness without demographics using social networks
David Liu, Virginie Do, Nicolas Usunier +1
Group fairness is a popular approach to prevent unfavorable treatment of individuals based on sensitive attributes such as race, gender, and disability. However, the reliance of gr…
Identifying and Mitigating Instability in Embeddings of the Degenerate Core
David Liu, Tina Eliassi-Rad
Are the embeddings of a graph's degenerate core stable? What happens to the embeddings of nodes in the degenerate core as we systematically remove periphery nodes (by repeated peel…
Improving realistic material property prediction using domain adaptation based machine learning
Jeffrey Hu, David Liu, Nihang Fu +1
Materials property prediction models are usually evaluated using random splitting of datasets into training and test datasets, which not only leads to over-estimated performance du…
Syntactic Complexity of Finite/Cofinite, Definite, and Reverse Definite Languages
Janusz Brzozowski, David Liu
We study the syntactic complexity of finite/cofinite, definite and reverse definite languages. The syntactic complexity of a class of languages is defined as the maximal size of sy…
How Low Can You Go? Practical cold-start performance limits in FaaS
Yue Tan, David Liu, Nanqinqin Li +1
Function-as-a-Service (FaaS) has recently emerged as a new cloud computing paradigm. It promises high utilization of data center resources through allocating resources on demand at…
Generating Synthetic Doctor-Patient Conversations for Long-form Audio Summarization
Yanis Labrak, David Grünert, Séverin Baroudi +11
Long-context audio reasoning is underserved in both training data and evaluation. Existing benchmarks target short-context tasks, and the open-ended generation tasks most relevant…
Universal Witnesses for State Complexity of Boolean Operations and Concatenation Combined with Star
Janusz Brzozowski, David Liu
We study the state complexity of boolean operations and product (concatenation, catenation) combined with star. We derive tight upper bounds for the symmetric differences and diffe…
Generative Sequential Notification Optimization via Multi-Objective Decision Transformers
Borja Ocejo, Ruofan Wang, Ke Liu +7
Notifications are an important communication channel for delivering timely and relevant information. Optimizing their delivery involves addressing complex sequential decision-makin…
Visual Search at Pinterest
Yushi Jing, David Liu, Dmitry Kislyuk +4
We demonstrate that, with the availability of distributed computation platforms such as Amazon Web Services and open-source tools, it is possible for a small engineering team to bu…
Coupling Artificial Neurons in BERT and Biological Neurons in the Human Brain
Xu Liu, Mengyue Zhou, Gaosheng Shi +6
Linking computational natural language processing (NLP) models and neural responses to language in the human brain on the one hand facilitates the effort towards disentangling the…
Symmetry, stability, and computation of degenerate lasing modes
David Liu, Bo Zhen, Li Ge +6
We present a general method to obtain the stable lasing solutions for the steady-state ab-initio lasing theory (SALT) for the case of a degenerate symmetric laser in two dimensions…
Feature-Align Network with Knowledge Distillation for Efficient Denoising
Lucas D. Young, Fitsum A. Reda, Rakesh Ranjan +6
We propose an efficient neural network for RAW image denoising. Although neural network-based denoising has been extensively studied for image restoration, little attention has bee…
APP-RUSS: Automated Path Planning for Robotic Ultrasound System
David Liu, Jerome Charton, Xiang Li +1
Autonomous robotic ultrasound System (RUSS) has been extensively studied. However, fully automated ultrasound image acquisition is still challenging, partly due to the lack of stud…
XTinyU-Net: Training-Free U-Net Scaling via Initialization-Time Sensitivity
Alvin Kimbowa, Moein Heidari, David Liu +1
While U-Net architectures remain the gold standard for medical image segmentation, their deployment in resource-constrained environments demands aggressive model compression. Howev…
Decoupled DMD: CFG Augmentation as the Spear, Distribution Matching as the Shield
Dongyang Liu, Peng Gao, David Liu +8
Diffusion model distillation has emerged as a powerful technique for creating efficient few-step and single-step generators. Among these, Distribution Matching Distillation (DMD) a…
Distribution Matching Distillation Meets Reinforcement Learning
Dengyang Jiang, Dongyang Liu, Zanyi Wang +12
Distribution Matching Distillation (DMD) facilitates efficient inference by distilling multi-step diffusion models into few-step variants. Concurrently, Reinforcement Learning (RL)…
Prompt and circumstance: A word-by-word LLM prompting approach to interlinear glossing for low-resource languages
Micha Elsner, David Liu
Partly automated creation of interlinear glossed text (IGT) has the potential to assist in linguistic documentation. We argue that LLMs can make this process more accessible to lin…