Source author record

Xiaoyu Wu

Xiaoyu Wu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Computer Vision cond-mat.mes-hall cond-mat.mtrl-sci Artificial Intelligence Computational Complexity Data Structures and Algorithms Machine Learning

Catalog footprint

What is connected

6works

7topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

Taming Outlier Tokens in Diffusion Transformers

We study outlier tokens in Diffusion Transformers (DiTs) for image generation. Prior work has shown that Vision Transformers (ViTs) can produce a small number of high-norm tokens that attract disproportionate attention while carrying limited local information, but their role in generative models remains underexplored. We show that this phenomenon appears in both the encoder and denoiser of modern Representation Autoencoder (RAE)-DiT pipelines: pretrained ViT encoders can produce outlier representations, and DiTs themselves can develop internal outlier tokens, especially in intermediate layers. Moreover, simply masking high-norm tokens does not improve performance, indicating that the problem is not only caused by a few extreme values, but is more closely related to corrupted local patch semantics. To address this issue, we introduce Dual-Stage Registers (DSR), a register-based intervention for both components: trained registers when available, recursive test-time registers otherwise, and diffusion registers for the denoiser. Across ImageNet and large-scale text-to-image generation, these interventions consistently reduce outlier artifacts and improve generation quality. Our results highlight outlier-token control as an important ingredient in building stronger DiTs.

preprint2022arXiv

Approximation Algorithms for Interdiction Problem with Packing Constraints

We study a bilevel optimization problem which is a zero-sum Stackelberg game. In this problem, there are two players, a leader and a follower, who pick items from a common set. Both the leader and the follower have their own (multi-dimensional) budgets, respectively. Each item is associated with a profit, which is the same to the leader and the follower, and will consume the leader's (follower's) budget if it is selected by the leader (follower). The leader and the follower will select items in a sequential way: First, the leader selects items within the leader's budget. Then the follower selects items from the remaining items within the follower's budget. The goal of the leader is to minimize the maximum profit that the follower can obtain. Let $s_A$ and $s_B$ be the dimension of the leader's and follower's budget, respectively. A special case of our problem is the bilevel knapsack problem studied by Caprara et al. [SIAM Journal on Optimization, 2014], where $s_A=s_B=1$. We consider the general problem and obtain an $(s_B+ε)$-approximation algorithm when $s_A$ and $s_B$ are both constant. In particular, if $s_B=1$, our algorithm implies a PTAS for the bilevel knapsack problem, which is the first O(1)-approximation algorithm. We also complement our result by showing that there does not exist any $(4/3-ε)$-approximation algorithm even if $s_A=1$ and $s_B=2$. We also consider a variant of our problem with resource augmentation when $s_A$ and $s_B$ are both part of the input. We obtain an O(1)-approximation algorithm with O(1)-resource augmentation, that is, we give an algorithm that returns a solution which exceeds the given leader's budget by O(1) times, and the objective value achieved by the solution is O(1) times the optimal objective value that respects the leader's budget.

preprint2022arXiv

Locality-aware Attention Network with Discriminative Dynamics Learning for Weakly Supervised Anomaly Detection

Video anomaly detection is recently formulated as a multiple instance learning task under weak supervision, in which each video is treated as a bag of snippets to be determined whether contains anomalies. Previous efforts mainly focus on the discrimination of the snippet itself without modeling the temporal dynamics, which refers to the variation of adjacent snippets. Therefore, we propose a Discriminative Dynamics Learning (DDL) method with two objective functions, i.e., dynamics ranking loss and dynamics alignment loss. The former aims to enlarge the score dynamics gap between positive and negative bags while the latter performs temporal alignment of the feature dynamics and score dynamics within the bag. Moreover, a Locality-aware Attention Network (LA-Net) is constructed to capture global correlations and re-calibrate the location preference across snippets, followed by a multilayer perceptron with causal convolution to obtain anomaly scores. Experimental results show that our method achieves significant improvements on two challenging benchmarks, i.e., UCF-Crime and XD-Violence.

preprint2020arXiv

Unexpected Giant Microwave Conductivity in a Nominally Silent BiFeO3 Domain Wall

Nanoelectronic devices based on ferroelectric domain walls (DWs), such as memories, transistors, and rectifiers, have been demonstrated in recent years. Practical high-speed electronics, on the other hand, usually demand operation frequencies in the giga-Hertz (GHz) regime, where the effect of dipolar oscillation is important. In this work, an unexpected giant GHz conductivity on the order of 103 S/m is observed in certain BiFeO3 DWs, which is about 100,000 times greater than the carrier-induced dc conductivity of the same walls. Surprisingly, the nominal configuration of the DWs precludes the ac conduction under an excitation electric field perpendicular to the surface. Theoretical analysis shows that the inclined DWs are stressed asymmetrically near the film surface, whereas the vertical walls in a control sample are not. The resultant imbalanced polarization profile can then couple to the out-of-plane microwave fields and induce power dissipation, which is confirmed by the phase-field modeling. Since the contributions from mobile-carrier conduction and bound-charge oscillation to the ac conductivity are equivalent in a microwave circuit, the research on local structural dynamics may open a new avenue to implement DW nano-devices for RF applications.

preprint2016arXiv

Uncovering edge states and electrical inhomogeneity in MoS2 field-effect transistors

The understanding of various types of disorders in atomically thin transition metal dichalcogenides (TMDs), including dangling bonds at the edges, chalcogen deficiencies in the bulk, and charges in the substrate, is of fundamental importance for their applications in electronics and photonics. Because of the imperfections, electrons moving on these two-dimensional (2D) crystals experience a spatially non-uniform Coulomb environment, whose effect on the charge transport has not been microscopically studied. Here, we report the mesoscopic conductance mapping in monolayer and few-layer MoS2 field-effect transistors (FETs) by microwave impedance microscopy (MIM). The spatial evolution of the insulator-to-metal transition is clearly resolved. Interestingly, as the transistors are gradually turned on, electrical conduction emerges initially at the edges before appearing in the bulk of MoS2 flakes, which can be explained by our first-principles calculations. The results unambiguously confirm that the contribution of edge states to the channel conductance is significant under the threshold voltage but negligible once the bulk of the TMD device becomes conductive. Strong conductance inhomogeneity, which is associated with the fluctuations of disorder potential in the 2D sheets, is also observed in the MIM images, providing a guideline for future improvement of the device performance.

preprint2015arXiv

Thickness-dependent Dielectric Constant of Few-layer In2Se3 Nano-flakes

The dielectric constant or relative permittivity of a dielectric material, which describes how the net electric field in the medium is reduced with respect to the external field, is a parameter of critical importance for charging and screening in electronic devices. Such a fundamental material property is intimately related to not only the polarizability of individual atoms, but also the specific atomic arrangement in the crystal lattice. In this letter, we present both experimental and theoretical investigations on the dielectric constant of few-layer In2Se3 nano-flakes grown on mica substrates by van der Waals epitaxy. A nondestructive microwave impedance microscope is employed to simultaneously quantify the number of layers and local electrical properties. The measured dielectric constant increases monotonically as a function of the thickness and saturates to the bulk value at around 6 ~ 8 quintuple layers. The same trend of layer-dependent dielectric constant is also revealed by first-principle calculations. Our results of the dielectric response, being ubiquitously applicable to layered 2D semiconductors, are expected to be significant for this vibrant research field.