Xinyi Zhao

Xinyi Zhao, Anna I. Thoma, Ralph Hertwig et al.

Women comprise the majority of students and early-career scholars in psychology, yet they are less likely to remain active in research over time. This pattern raises a central question: At what stages of academic careers do women disproportionately leave academia, and what factors drive their attrition? Using large-scale bibliometric data tracking 78,216 psychologists who began publishing between 2000 and 2014, we examine gender differences in research career attrition operationalized through publishing activity across the full trajectory from entry onward. Although women accounted for more than 60\% of new entrants, they experienced higher attrition rates than men, with the gender gap peaking approximately five years after first publication. Early-career performance, particularly first-authored publications, was the strongest predictor of subsequent retention, whereas last-authored publications were most closely associated with continued activity at later career stages. Collaboration patterns and institutional context also shaped career persistence, though less strongly than publication indicators. Notably, gender differences in research attrition persisted even after accounting for these career determinants, especially during early career stages. These findings suggest that gender inequality in psychology is driven less by recruitment than by differential retention over time. Addressing early-career vulnerability may therefore be essential to achieving equitable representation in senior academic leadership within the discipline.

Xinyi Zhao, Weixin Liang, James Zou · stanford

Data is the fuel powering AI and creates tremendous value for many domains. However, collecting datasets for AI is a time-consuming, expensive, and complicated endeavor. For practitioners, data investment remains to be a leap of faith in practice. In this work, we study the data budgeting problem and formulate it as two sub-problems: predicting (1) what is the saturating performance if given enough data, and (2) how many data points are needed to reach near the saturating performance. Different from traditional dataset-independent methods like PowerLaw, we proposed a learning method to solve data budgeting problems. To support and systematically evaluate the learning-based method for data budgeting, we curate a large collection of 383 tabular ML datasets, along with their data vs performance curves. Our empirical evaluation shows that it is possible to perform data budgeting given a small pilot study dataset with as few as $50$ data points.

Shengxi Li, Xinyi Zhao, Ljubisa Stankovic et al.

The success of convolution neural networks (CNN) has been revolutionising the way we approach and use intelligent machines in the Big Data era. Despite success, CNNs have been consistently put under scrutiny owing to their \textit{black-box} nature, an \textit{ad hoc} manner of their construction, together with the lack of theoretical support and physical meanings of their operation. This has been prohibitive to both the quantitative and qualitative understanding of CNNs, and their application in more sensitive areas such as AI for health. We set out to address these issues, and in this way demystify the operation of CNNs, by employing the perspective of matched filtering. We first illuminate that the convolution operation, the very core of CNNs, represents a matched filter which aims to identify the presence of features in input data. This then serves as a vehicle to interpret the convolution-activation-pooling chain in CNNs under the theoretical umbrella of matched filtering, a common operation in signal processing. We further provide extensive examples and experiments to illustrate this connection, whereby the learning in CNNs is shown to also perform matched filtering, which further sheds light onto physical meaning of learnt parameters and layers. It is our hope that this material will provide new insights into the understanding, constructing and analysing of CNNs, as well as paving the way for developing new methods and architectures of CNNs.

Simran Arora, Aman Timalsina, Aaryan Singhal et al.

Recurrent large language models that compete with Transformers in language modeling perplexity are emerging at a rapid rate (e.g., Mamba, RWKV). Excitingly, these architectures use a constant amount of memory during inference. However, due to the limited memory, recurrent LMs cannot recall and use all the information in long contexts leading to brittle in-context learning (ICL) quality. A key challenge for efficient LMs is selecting what information to store versus discard. In this work, we observe the order in which information is shown to the LM impacts the selection difficulty. To formalize this, we show that the hardness of information recall reduces to the hardness of a problem called set disjointness (SD), a quintessential problem in communication complexity that requires a streaming algorithm (e.g., recurrent model) to decide whether inputted sets are disjoint. We empirically and theoretically show that the recurrent memory required to solve SD changes with set order, i.e., whether the smaller set appears first in-context. Our analysis suggests, to mitigate the reliance on data order, we can put information in the right order in-context or process prompts non-causally. Towards that end, we propose: (1) JRT-Prompt, where context gets repeated multiple times in the prompt, effectively showing the model all data orders. This gives $11.0 \pm 1.3$ points of improvement, averaged across $16$ recurrent LMs and the $6$ ICL tasks, with $11.9\times$ higher throughput than FlashAttention-2 for generation prefill (length $32$k, batch size $16$, NVidia H100). We then propose (2) JRT-RNN, which uses non-causal prefix-linear-attention to process prompts and provides $99\%$ of Transformer quality at $360$M params., $30$B tokens and $96\%$ at $1.3$B params., $50$B tokens on average across the tasks, with $19.2\times$ higher throughput for prefill than FA2.

Xinyi Zhao, Prasanna Raut, Chaoyue Zhao et al.

Wildfires pose an increasing threat to the safety and reliability of power systems, particularly in distribution networks located in fire-prone regions. To mitigate ignition risk from electrical infrastructure, utilities often employ safety power shutoffs, which proactively de-energize high-risk lines during hazardous weather and restore them once conditions improve. While this strategy can result in temporary load loss, it helps prevent equipment damage and wildfire ignition development in the system. In this paper, we develop a state-based decision-making framework to optimize such switching actions over time, with the goal of minimizing total operational costs throughout a wildfire event. The model represents network topologies as Markov states, with transitions influenced by both exogenous weather conditions and endogenous power flow dynamics. To address the computational challenges posed by the large state and action spaces, we propose an approximate dynamic programming algorithm based on post-decision states. The effectiveness and scalability of the proposed approach are demonstrated through case studies on 54-bus and 138-bus distribution systems, showcasing its potential for enhancing wildfire resilience across different grid configurations.

Xinyi Zhao, Congjing Zhang, Pei Guo et al.

Video anomaly detection (VAD) is essential for enhancing safety and security by identifying unusual events across different environments. Existing VAD benchmarks, however, are primarily designed for general-purpose scenarios, neglecting the specific characteristics of smart home applications. To bridge this gap, we introduce SmartHome-Bench, the first comprehensive benchmark specially designed for evaluating VAD in smart home scenarios, focusing on the capabilities of multi-modal large language models (MLLMs). Our newly proposed benchmark consists of 1,203 videos recorded by smart home cameras, organized according to a novel anomaly taxonomy that includes seven categories, such as Wildlife, Senior Care, and Baby Monitoring. Each video is meticulously annotated with anomaly tags, detailed descriptions, and reasoning. We further investigate adaptation methods for MLLMs in VAD, assessing state-of-the-art closed-source and open-source models with various prompting techniques. Results reveal significant limitations in the current models' ability to detect video anomalies accurately. To address these limitations, we introduce the Taxonomy-Driven Reflective LLM Chain (TRLC), a new LLM chaining framework that achieves a notable 11.62% improvement in detection accuracy. The benchmark dataset and code are publicly available at https://github.com/Xinyi-0724/SmartHome-Bench-LLM.

Kevin Mayer, Alex Vesel, Xinyi Zhao et al.

3D building models are critical for applications in architecture, energy simulation, and navigation. Yet, generating accurate and semantically rich 3D buildings automatically remains a major challenge due to the lack of large-scale annotated datasets in the public domain. Inspired by the success of synthetic data in computer vision, we introduce SYNBUILD-3D, a large, diverse, and multi-modal dataset of over 6.2 million synthetic 3D residential buildings at Level of Detail (LoD) 4. In the dataset, each building is represented through three distinct modalities: a semantically enriched 3D wireframe graph at LoD 4 (Modality I), the corresponding floor plan images (Modality II), and a LiDAR-like roof point cloud (Modality III). The semantic annotations for each building wireframe are derived from the corresponding floor plan images and include information on rooms, doors, and windows. Through its tri-modal nature, future work can use SYNBUILD-3D to develop novel generative AI algorithms that automate the creation of 3D building models at LoD 4, subject to predefined floor plan layouts and roof geometries, while enforcing semantic-geometric consistency. Dataset and code samples are publicly available at https://github.com/kdmayer/SYNBUILD-3D.

Xin Liu, Ziqing Xie, Yongjun Yuan et al.

This paper investigates the dynamics of spin-2 Bose-Einstein condensates (BECs) with rotation and spin-orbit coupling (SOC). In order to better simulate the dynamics, we present an efficient high-order compact splitting Fourier spectral method. This method splits the Hamiltonian into a linear part, which consists of the Laplace, rotation and SOC terms, and a nonlinear part that includes all the remaining terms. The wave function is well approximated by the Fourier spectral method and is numerically accessed with discrete Fast Fourier transform (FFT). For linear subproblem, the handling of rotation term and SOC term poses a major challenge. Using a function mapping based on rotation, we can integrate the linear subproblem exactly and explicitly. This mapping we propose not only helps eliminate the rotation term, but also prevents the SOC term from evolving into a time-dependent form. The nonlinear subproblem is integrated analytically in physical space. Such "compact" splitting involves only two operators and facilitates the design of high-order splitting schemes. Our method is spectrally accurate in space and high order in time. It is efficient, explicit, unconditionally stable and simple to implement. In addition, we derive some dynamical properties and carry out a systematic study, including accuracy and efficiency tests, dynamical property verification, the SOC effects and dynamics of vortex lattice.

Yu Zhang, Xinyi Zhao, Chongke Bi et al.

Semantic segmentation of 3D point clouds is important for many applications, such as autonomous driving. To train semantic segmentation models, labeled point cloud segmentation datasets are essential. Meanwhile, point cloud labeling is time-consuming for annotators, which typically involves tuning the camera viewpoint and selecting points by lasso. To reduce the time cost of point cloud labeling, we propose a viewpoint recommendation approach to reduce annotators' labeling time costs. We adapt Fitts' law to model the time cost of lasso selection in point clouds. Using the modeled time cost, the viewpoint that minimizes the lasso selection time cost is recommended to the annotator. We build a data labeling system for semantic segmentation of 3D point clouds that integrates our viewpoint recommendation approach. The system enables users to navigate to recommended viewpoints for efficient annotation. Through an ablation study, we observed that our approach effectively reduced the data labeling time cost. We also qualitatively compare our approach with previous viewpoint selection approaches on different datasets.

Xinyi Zhao, Samin Aref, Emilio Zagheni et al.

The international migration of researchers is an important dimension of scientific mobility, and has been the subject of considerable policy debate. However, tracking the migration life courses of researchers is challenging due to data limitations. In this study, we use Scopus bibliometric data on eight million publications from 1.1 million researchers who have published at least once with an affiliation address from Germany in 1996-2020. We construct the partial life histories of published researchers in this period and explore both their out-migration and the subsequent return of a subset of this group: the returnees. Our analyses shed light on the career stages and gender disparities between researchers who remain in Germany, those who emigrate, and those who eventually return. We find that the return migration streams are even more gender imbalanced, which points to the need for additional efforts to encourage female researchers to come back to Germany. We document a slightly declining trend in return migration among more recent cohorts of researchers who left Germany, which, for most disciplines, was associated with a decrease in the German collaborative ties of these researchers. Moreover, we find that the gender disparities for the most gender imbalanced disciplines are unlikely to be mitigated by return migration given the gender compositions of the cohorts of researchers who have left Germany and of those who have returned. This analysis uncovers new dimensions of migration among scholars by investigating the return migration of published researchers, which is critical for the development of science policy.

Chaojian Yu, Xinyi Zhao, Qi Zheng et al.

Fine-grained visual recognition is challenging because it highly relies on the modeling of various semantic parts and fine-grained feature learning. Bilinear pooling based models have been shown to be effective at fine-grained recognition, while most previous approaches neglect the fact that inter-layer part feature interaction and fine-grained feature learning are mutually correlated and can reinforce each other. In this paper, we present a novel model to address these issues. First, a cross-layer bilinear pooling approach is proposed to capture the inter-layer part feature relations, which results in superior performance compared with other bilinear pooling based approaches. Second, we propose a novel hierarchical bilinear pooling framework to integrate multiple cross-layer bilinear features to enhance their representation capability. Our formulation is intuitive, efficient and achieves state-of-the-art results on the widely used fine-grained recognition datasets.