arXiv subjects

Wenxuan Li

Publications and source records attributed to Wenxuan Li.

52 records · Page 3Linked to original sources

arXiv2025

Navigating in High-Dimensional Search Space: A Hierarchical Bayesian Optimization Approach

Optimizing black-box functions in high-dimensional search spaces has been known to be challenging for traditional Bayesian Optimization (BO). In this paper, we introduce HiBO, a novel hierarchical algorithm integrating global-level search space partitioning information into the acquisition strategy of a local BO-based optimizer. HiBO employs a search-tree-based global-level navigator to adaptively split the search space into partitions with different sampling potential. The local optimizer then utilizes this global-level information to guide its acquisition strategy towards most promising regions within the search space. A comprehensive set of evaluations demonstrates that HiBO outperforms state-of-the-art methods in high-dimensional synthetic benchmarks and presents significant practical effectiveness in the real-world task of tuning configurations of database management systems (DBMSs).

cs.LG↗

arXiv2025

ClueCart: Supporting Game Story Interpretation and Narrative Inference from Fragmented Clues

Indexical storytelling is gaining popularity in video games, where the narrative unfolds through fragmented clues. This approach fosters player-generated content and discussion, as story interpreters piece together the overarching narrative from these scattered elements. However, the fragmented and non-linear nature of the clues makes systematic categorization and interpretation challenging, potentially hindering efficient story reconstruction and creative engagement. To address these challenges, we first proposed a hierarchical taxonomy to categorize narrative clues, informed by a formative study. Using this taxonomy, we designed ClueCart, a creativity support tool aimed at enhancing creators' ability to organize story clues and facilitate intricate story interpretation. We evaluated ClueCart through a between-subjects study (N=40), using Miro as a baseline. The results showed that ClueCart significantly improved creators' efficiency in organizing and retrieving clues, thereby better supporting their creative processes. Additionally, we offer design insights for future studies focused on player-centric narrative analysis.

cs.HC↗

arXiv2025

Touchstone Benchmark: Are We on the Right Way for Evaluating AI Algorithms for Medical Segmentation?

How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks--which, differing from algorithms, are more flexible and can support different algorithms--including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.

cs.CV↗

arXiv2025

How Well Do Supervised 3D Models Transfer to Medical Imaging Tasks?

The pre-training and fine-tuning paradigm has become prominent in transfer learning. For example, if the model is pre-trained on ImageNet and then fine-tuned to PASCAL, it can significantly outperform that trained on PASCAL from scratch. While ImageNet pre-training has shown enormous success, it is formed in 2D, and the learned features are for classification tasks; when transferring to more diverse tasks, like 3D image segmentation, its performance is inevitably compromised due to the deviation from the original ImageNet context. A significant challenge lies in the lack of large, annotated 3D datasets rivaling the scale of ImageNet for model pre-training. To overcome this challenge, we make two contributions. Firstly, we construct AbdomenAtlas 1.1 that comprises 9,262 three-dimensional computed tomography (CT) volumes with high-quality, per-voxel annotations of 25 anatomical structures and pseudo annotations of seven tumor types. Secondly, we develop a suite of models that are pre-trained on our AbdomenAtlas 1.1 for transfer learning. Our preliminary analyses indicate that the model trained only with 21 CT volumes, 672 masks, and 40 GPU hours has a transfer learning ability similar to the model trained with 5,050 (unlabeled) CT volumes and 1,152 GPU hours. More importantly, the transfer learning ability of supervised models can further scale up with larger annotated datasets, achieving significantly better performance than preexisting pre-trained models, irrespective of their pre-training methodologies or data sources. We hope this study can facilitate collective efforts in constructing larger 3D medical datasets and more releases of supervised pre-trained models.

eess.IV↗

arXiv2024

Co-Fix3D: Enhancing 3D Object Detection with Collaborative Refinement

3D object detection in driving scenarios faces the challenge of complex road environments, which can lead to the loss or incompleteness of key features, thereby affecting perception performance. To address this issue, we propose an advanced detection framework called Co-Fix3D. Co-Fix3D integrates Local and Global Enhancement (LGE) modules to refine Bird's Eye View (BEV) features. The LGE module uses Discrete Wavelet Transform (DWT) for pixel-level local optimization and incorporates an attention mechanism for global optimization. To handle varying detection difficulties, we adopt multi-head LGE modules, enabling each module to focus on targets with different levels of detection complexity, thus further enhancing overall perception capability. Experimental results show that on the nuScenes dataset's LiDAR benchmark, Co-Fix3D achieves 69.4\% mAP and 73.5\% NDS, while on the multimodal benchmark, it achieves 72.3\% mAP and 74.7\% NDS. The source code is publicly available at \href{https://github.com/rubbish001/Co-Fix3d}{https://github.com/rubbish001/Co-Fix3d}.

cs.CV↗

arXiv2024

Label Critic: Design Data Before Models

As medical datasets rapidly expand, creating detailed annotations of different body structures becomes increasingly expensive and time-consuming. We consider that requesting radiologists to create detailed annotations is unnecessarily burdensome and that pre-existing AI models can largely automate this process. Following the spirit don't use a sledgehammer on a nut, we find that, rather than creating annotations from scratch, radiologists only have to review and edit errors if the Best-AI Labels have mistakes. To obtain the Best-AI Labels among multiple AI Labels, we developed an automatic tool, called Label Critic, that can assess label quality through tireless pairwise comparisons. Extensive experiments demonstrate that, when incorporated with our developed Image-Prompt pairs, pre-existing Large Vision-Language Models (LVLM), trained on natural images and texts, achieve 96.5% accuracy when choosing the best label in a pair-wise comparison, without extra fine-tuning. By transforming the manual annotation task (30-60 min/scan) into an automatic comparison task (15 sec/scan), we effectively reduce the manual efforts required from radiologists by an order of magnitude. When the Best-AI Labels are sufficiently accurate (81% depending on body structures), they will be directly adopted as the gold-standard annotations for the dataset, with lower-quality AI Labels automatically discarded. Label Critic can also check the label quality of a single AI Label with 71.8% accuracy when no alternatives are available for comparison, prompting radiologists to review and edit if the estimated quality is low (19% depending on body structures).

cs.CV↗

arXiv2024

Neural Dehydration: Effective Erasure of Black-box Watermarks from DNNs with Limited Data

To protect the intellectual property of well-trained deep neural networks (DNNs), black-box watermarks, which are embedded into the prediction behavior of DNN models on a set of specially-crafted samples and extracted from suspect models using only API access, have gained increasing popularity in both academy and industry. Watermark robustness is usually implemented against attackers who steal the protected model and obfuscate its parameters for watermark removal. However, current robustness evaluations are primarily performed under moderate attacks or unrealistic settings. Existing removal attacks could only crack a small subset of the mainstream black-box watermarks, and fall short in four key aspects: incomplete removal, reliance on prior knowledge of the watermark, performance degradation, and high dependency on data. In this paper, we propose a watermark-agnostic removal attack called \textsc{Neural Dehydration} (\textit{abbrev.} \textsc{Dehydra}), which effectively erases all ten mainstream black-box watermarks from DNNs, with only limited or even no data dependence. In general, our attack pipeline exploits the internals of the protected model to recover and unlearn the watermark message. We further design target class detection and recovered sample splitting algorithms to reduce the utility loss and achieve data-free watermark removal on five of the watermarking schemes. We conduct comprehensive evaluation of \textsc{Dehydra} against ten mainstream black-box watermarks on three benchmark datasets and DNN architectures. Compared with existing removal attacks, \textsc{Dehydra} achieves strong removal effectiveness across all the covered watermarks, preserving at least $90\%$ of the stolen model utility, under the data-limited settings, i.e., less than $2\%$ of the training data or even data-free.

cs.CR↗

arXiv2024

A Semantic Space is Worth 256 Language Descriptions: Make Stronger Segmentation Models with Descriptive Properties

This paper introduces ProLab, a novel approach using property-level label space for creating strong interpretable segmentation models. Instead of relying solely on category-specific annotations, ProLab uses descriptive properties grounded in common sense knowledge for supervising segmentation models. It is based on two core designs. First, we employ Large Language Models (LLMs) and carefully crafted prompts to generate descriptions of all involved categories that carry meaningful common sense knowledge and follow a structured format. Second, we introduce a description embedding model preserving semantic correlation across descriptions and then cluster them into a set of descriptive properties (e.g., 256) using K-Means. These properties are based on interpretable common sense knowledge consistent with theories of human recognition. We empirically show that our approach makes segmentation models perform stronger on five classic benchmarks (e.g., ADE20K, COCO-Stuff, Pascal Context, Cityscapes, and BDD). Our method also shows better scalability with extended training steps than category-level supervision. Our interpretable segmentation framework also emerges with the generalization ability to segment out-of-domain or unknown categories using only in-domain descriptive properties. Code is available at https://github.com/lambert-x/ProLab.

cs.CV↗

arXiv2024

InterGen: Diffusion-based Multi-human Motion Generation under Complex Interactions

We have recently seen tremendous progress in diffusion advances for generating realistic human motions. Yet, they largely disregard the multi-human interactions. In this paper, we present InterGen, an effective diffusion-based approach that incorporates human-to-human interactions into the motion diffusion process, which enables layman users to customize high-quality two-person interaction motions, with only text guidance. We first contribute a multimodal dataset, named InterHuman. It consists of about 107M frames for diverse two-person interactions, with accurate skeletal motions and 23,337 natural language descriptions. For the algorithm side, we carefully tailor the motion diffusion model to our two-person interaction setting. To handle the symmetry of human identities during interactions, we propose two cooperative transformer-based denoisers that explicitly share weights, with a mutual attention mechanism to further connect the two denoising processes. Then, we propose a novel representation for motion input in our interaction diffusion model, which explicitly formulates the global relations between the two performers in the world frame. We further introduce two novel regularization terms to encode spatial relations, equipped with a corresponding damping scheme during the training of our interaction diffusion model. Extensive experiments validate the effectiveness and generalizability of InterGen. Notably, it can generate more diverse and compelling two-person motions than previous methods and enables various downstream applications for human interactions.

cs.CV↗

arXiv2023

MedShapeNet -- A Large-Scale Dataset of 3D Medical Shapes for Computer Vision

Prior to the deep learning era, shape was commonly used to describe the objects. Nowadays, state-of-the-art (SOTA) algorithms in medical imaging are predominantly diverging from computer vision, where voxel grids, meshes, point clouds, and implicit surface models are used. This is seen from numerous shape-related publications in premier vision conferences as well as the growing popularity of ShapeNet (about 51,300 models) and Princeton ModelNet (127,915 models). For the medical domain, we present a large collection of anatomical shapes (e.g., bones, organs, vessels) and 3D models of surgical instrument, called MedShapeNet, created to facilitate the translation of data-driven vision algorithms to medical applications and to adapt SOTA vision algorithms to medical problems. As a unique feature, we directly model the majority of shapes on the imaging data of real patients. As of today, MedShapeNet includes 23 dataset with more than 100,000 shapes that are paired with annotations (ground truth). Our data is freely accessible via a web interface and a Python application programming interface (API) and can be used for discriminative, reconstructive, and variational benchmarks as well as various applications in virtual, augmented, or mixed reality, and 3D printing. Exemplary, we present use cases in the fields of classification of brain tumors, facial and skull reconstructions, multi-class anatomy completion, education, and 3D printing. In future, we will extend the data and improve the interfaces. The project pages are: https://medshapenet.ikim.nrw/ and https://github.com/Jianningli/medshapenet-feedback

cs.CV↗

arXiv2022

Rigidity of the Stochastic Airy Operator

We prove that the spectrum of the stochastic Airy operator is rigid in the sense of Ghosh and Peres (Duke Math. J., 166(10):1789--1858, 2017) for Dirichlet and Robin boundary conditions. This proves the rigidity of the Airy-$β$ point process and the soft-edge limit of rank-$1$ perturbations of Gaussian $β$-Ensembles for any $β>0$, and solves an open problem mentioned in a previous work of Bufetov, Nikitin, and Qiu (Mosc. Math. J., 19(2):217--274, 2019). Our proof uses a combination of the semigroup theory of the stochastic Airy operator and the techniques for studying insertion and deletion tolerance of point processes developed by Holroyd and Soo (Electron. J. Probab., 18:no. 74, 24, 2013).

math.PR↗

arXiv2021

Boosting the Capability of Intelligent Vulnerability Detection by Training in a Human-Learning Manner

Due to its powerful automatic feature extraction, deep learning (DL) has been widely used in source code vulnerability detection. However, although it performs well on artificial datasets, its performance is not satisfactory when detecting real-world vulnerabilities due to the high complexity of real-world samples. In this paper, we propose to train DL-based vulnerability detection models in a human-learning manner, that is, start with the simplest samples and then gradually transition to difficult knowledge. Specifically, we design a novel framework (Humer) that can enhance the detection ability of DL-based vulnerability detectors. To validate the effectiveness of Humer, we select five state-of-the-art DL-based vulnerability detection models (TokenCNN, VulDeePecker, StatementGRU, ASTGRU, and Devign) to complete our evaluations. Through the results, we find that the use of Humer can increase the F1 of these models by an average of 10.5%. Moreover, Humer can make the model detect up to 16.7% more real-world vulnerabilities. Meanwhile, we also conduct a case study to uncover vulnerabilities from real-world open source products by using these enhanced DL-based vulnerability detectors. Through the results, we finally discover 281 unreported vulnerabilities in NVD, of which 98 have been silently patched by vendors in the latest version of corresponding products, but 159 still exist in the products.

cs.CR↗

arXiv2021

Consequences of the packing problem

We study several consequences of the packing problem, a conjecture from combinatorial optimization, using algebraic invariants of square-free monomial ideals. While the packing problem is currently unresolved, we successfully settle the validity of its consequences. Our work prompts additional questions and conjectures, which are presented together with their motivation.

math.AC↗

arXiv2017

The Role of Weak Interactions in Characterizing Peptide Folding Preferences using a QTAIM Interpretation of the Ramachandran Plot (ϕ-ψ)

The Ramachandran plot is a potent way to understand structures of biomolecules, however, the original formulation of the Ramachandran plot only considers backbone conformations. We formulate a new interpretation of the original Ramachandran plot ($ϕ-ψ$) that can include a description of the weaker interactions including both the hydrogen bonds and H$---$H bonds as a new way to derive insights into the phenomenon of peptide folding. We use QTAIM (quantum theory of atoms in molecules) to interpret the Ramachandran plot. Specifically, we show that QTAIM analysis permits identifying key regions of the Ramachandran plot without the need for massive data sets. A highly non-linear relationship is found between the QTAIM vector-derived interpreted Ramachandran plot and the conventional Ramachandran plot ($ϕ-ψ$) demonstrating that this new approach is not a trivial coordinate transformation. An investigation of both the backbone and the weaker bonds within the framework of the QTAIM interpreted Ramachandran plot was found to be in line with physical intuition. The least-preferred directions calculated for the hydrogen bonds and H$---$H bonds were found to coincide with the 'unlikely' regions of the Ramachandran plot.

physics.chem-ph↗

arXiv2016

lfda: An R Package for Local Fisher Discriminant Analysis and Visualization

Local Fisher discriminant analysis is a localized variant of Fisher discriminant analysis and it is popular for supervised dimensionality reduction method. lfda is an R package for performing local Fisher discriminant analysis, including its variants such as kernel local Fisher discriminant analysis and semi-supervised local Fisher discriminant analysis. It also provides visualization functions to easily visualize the dimension reduction results by using either rgl for 3D visualization or ggfortify for 2D visualization in ggplot2 style.

stat.AP↗

arXiv2016

Comparative Density Functional Theory and Density Functional Tight Binding Study of Arginine and Arginine-Rich Cell-Penetrating Peptide TAT Adsorption on Anatase TiO2

We present a comparative Density Functional Theory (DFT) and Density Functional Tight Binding (DFTB) study of geometries and electronic structures of arginine (Arg), arginine adsorbed on the anatase (101) surface of titania in several adsorption configurations, and of an arginine-rich cell penetrating peptide TAT and its adsorption on the surface of TiO2. Two DFTB parameterizations are considered, tiorg-0-1/mio-1-1 and matsci-0-3. While there is good agreement in the structures and relative energies of Arg and peptide conformers between DFT and DFTB, both adsorption geometries and energies are noticeably different for Arg adsorbed on TiO2. The tiorg-0-1/mio-1-1 parameterization performs better than matsci-0-3. We relate this difference to the difference in electronic structures resulting from the two methods (DFT and DFTB) and specifically to the band alignment between the molecule and the oxide. We show that the band alignment of TAT and of TiO2 modeled with DFTB is qualitatively correct but that with DFT using the PBE functional is not. This is specific to the modeling of large molecules where the HOMO is close to the conduction band of the oxide. We therefore report a case where the approximate DFT-based method - DFTB (with which the correct band structure can be effectively obtained) - performs better than the DFT itself with a functional approximation feasible for the modeling of large bio-inorganic interfaces, i.e. GGA (as opposed to hybrid functionals which are impractical at such a scale). Our results highlight the utility of the DFTB method for the modeling of bioinorganic interfaces not only from the CPU cost perspective but also from the accuracy point of view.

cond-mat.mtrl-sci↗