Search

The development of person search techniques has been greatly promoted in recent years for its superior practicality and challenging goals. Despite their significant progress, existing person search models still lack the ability to continually learn from increaseing real-world data and adaptively process input from different domains. To this end, this work introduces the continual person search task that sequentially learns on multiple domains and then performs person search on all seen domains. This requires balancing the stability and plasticity of the model to continually learn new knowledge without catastrophic forgetting. For this, we propose a Prompt-based Continual Person Search (PoPS) model in this paper. First, we design a compositional person search transformer to construct an effective pre-trained transformer without exhaustive pre-training from scratch on large-scale person search data. This serves as the fundamental for prompt-based continual learning. On top of that, we design a domain incremental prompt pool with a diverse attribute matching module. For each domain, we independently learn a set of prompts to encode the domain-oriented knowledge. Meanwhile, we jointly learn a group of diverse attribute projections and prototype embeddings to capture discriminative domain attributes. By matching an input image with the learned attributes across domains, the learned prompts can be properly selected for model inference. Extensive experiments are conducted to validate the proposed method for continual person search. The source code is available at https://github.com/PatrickZad/PoPS., Comment: ACM MM 2024

The emergence of Large Language Models (LLMs) has necessitated the adoption of parallel training techniques, involving the deployment of thousands of GPUs to train a single model. Unfortunately, we have found that the efficiency of current parallel training is often suboptimal, largely due to the following two main issues. Firstly, hardware failures are inevitable, leading to interruptions in the training tasks. The inability to quickly identify the faulty components results in a substantial waste of GPU resources. Secondly, since GPUs must wait for parameter synchronization to complete before proceeding to the next round of computation, network congestions can greatly increase the waiting time for GPUs. To address these challenges, this paper introduces a communication-driven solution, namely the C4. The key insights of C4 are two folds. First, in parallel training, collective communication exhibits periodic and homogeneous characteristics, so any anomalies are certainly due to some form of hardware malfunction. By leveraging this feature, C4 can rapidly identify the faulty components, swiftly isolate the anomaly, and restart the task, thereby avoiding resource wastage caused by delays in anomaly detection. Second, the predictable communication model of collective communication, involving few large flows, allows C4 to efficiently execute traffic planning, substantially reducing network congestion. C4 has been extensively implemented across our production systems, cutting error-induced overhead by roughly 30% and enhancing runtime performance by about 15% for certain applications with moderate communication costs.

We prove two results on converse theorems for Hilbert modular forms over totally real fields of degree $r>1$. The first result recovers a Hilbert modular form (of some level) from an $L$-series satisfying functional equations twisted by all the unramified Hecke characters. The second result assumes both the 'unramified' functional equations and an Euler product, and recovers a Hilbert modular form of the expected level predicted by the shape of the functional equations. Our result generalizes the current converse theorems for $\mathrm{GL}_2$ in the case of Hilbert modular forms in that we completely remove the assumptions on ramified twists., Comment: 23 pages

The dependence of Natural Language Processing (NLP) intelligent software on Large Language Models (LLMs) is increasingly prominent, underscoring the necessity for robustness testing. Current testing methods focus solely on the robustness of LLM-based software to prompts. Given the complexity and diversity of real-world inputs, studying the robustness of LLMbased software in handling comprehensive inputs (including prompts and examples) is crucial for a thorough understanding of its performance. To this end, this paper introduces RITFIS, a Robust Input Testing Framework for LLM-based Intelligent Software. To our knowledge, RITFIS is the first framework designed to assess the robustness of LLM-based intelligent software against natural language inputs. This framework, based on given threat models and prompts, primarily defines the testing process as a combinatorial optimization problem. Successful test cases are determined by a goal function, creating a transformation space for the original examples through perturbation means, and employing a series of search methods to filter cases that meet both the testing objectives and language constraints. RITFIS, with its modular design, offers a comprehensive method for evaluating the robustness of LLMbased intelligent software. RITFIS adapts 17 automated testing methods, originally designed for Deep Neural Network (DNN)-based intelligent software, to the LLM-based software testing scenario. It demonstrates the effectiveness of RITFIS in evaluating LLM-based intelligent software through empirical validation. However, existing methods generally have limitations, especially when dealing with lengthy texts and structurally complex threat models. Therefore, we conducted a comprehensive analysis based on five metrics and provided insightful testing method optimization strategies, benefiting both researchers and everyday users.

Scattering-based single-particle tracking (S-SPT) has opened new avenues for highly sensitive label-free detection and characterization of nanoscopic objects, making it particularly attractive for various analytical applications. However, a long-standing issue hindering its widespread applicability is its high technical demands on optical systems. The most promising solution entails implementing on-chip light-scattering enhancement, but the existing field-enhancement technology fails as their highly localized field is insufficient to cover the three-dimensional trajectory of particles within the interrogation time. Here, we present a straightforward and robust on-chip microlens-based strategy for light-scattering enhancement, providing an enhancement range ten times greater than that of near-field optical techniques. These properties are attributed to the increased long-range optical fields and complex composite interactions between two closely spaced structures. Thanks to this strategy, we demonstrate that high-performance S-SPT can be achieved, for the first time, under a conventional bright-field microscope with illumination powers over 1,000 times lower than typically required. This significantly reduces the technical demands of S-SPT, representing a significant step forward in facilitating its practical application in biophotonics, biosensors, diagnostics, and other fields., Comment: 29 pages,4 figures

Salient object detection (SOD) in panoramic video is still in the initial exploration stage. The indirect application of 2D video SOD method to the detection of salient objects in panoramic video has many unmet challenges, such as low detection accuracy, high model complexity, and poor generalization performance. To overcome these hurdles, we design an Inter-Layer Attention (ILA) module, an Inter-Layer weight (ILW) module, and a Bi-Modal Attention (BMA) module. Based on these modules, we propose a Spatial-Temporal Dual-Mode Mixed Flow Network (STDMMF-Net) that exploits the spatial flow of panoramic video and the corresponding optical flow for SOD. First, the ILA module calculates the attention between adjacent level features of consecutive frames of panoramic video to improve the accuracy of extracting salient object features from the spatial flow. Then, the ILW module quantifies the salient object information contained in the features of each level to improve the fusion efficiency of the features of each level in the mixed flow. Finally, the BMA module improves the detection accuracy of STDMMF-Net. A large number of subjective and objective experimental results testify that the proposed method demonstrates better detection accuracy than the state-of-the-art (SOTA) methods. Moreover, the comprehensive performance of the proposed method is better in terms of memory required for model inference, testing time, complexity, and generalization performance.

The rapid adoption of AI and convenience offered by cloud services have resulted in the growing demands for GPUs in the cloud. Generally, GPUs are physically attached to host servers as PCIe devices. However, the fixed assembly combination of host servers and GPUs is extremely inefficient in resource utilization, upgrade, and maintenance. Due to these issues, the GPU disaggregation technique has been proposed to decouple GPUs from host servers. It aggregates GPUs into a pool, and allocates GPU node(s) according to user demands. However, existing GPU disaggregation systems have flaws in software-hardware compatibility, disaggregation scope, and capacity. In this paper, we present a new implementation of datacenter-scale GPU disaggregation, named DxPU. DxPU efficiently solves the above problems and can flexibly allocate as many GPU node(s) as users demand. In order to understand the performance overhead incurred by DxPU, we build up a performance model for AI specific workloads. With the guidance of modeling results, we develop a prototype system, which has been deployed into the datacenter of a leading cloud provider for a test run. We also conduct detailed experiments to evaluate the performance overhead caused by our system. The results show that the overhead of DxPU is less than 10%, compared with native GPU servers, in most of user scenarios., Comment: 23 pages, 6 figures, published in ACM Transactions on Architecture and Code Optimization

End-to-end person search aims to jointly detect and re-identify a target person in raw scene images with a unified model. The detection task unifies all persons while the re-id task discriminates different identities, resulting in conflict optimal objectives. Existing works proposed to decouple end-to-end person search to alleviate such conflict. Yet these methods are still sub-optimal on one or two of the sub-tasks due to their partially decoupled models, which limits the overall person search performance. In this paper, we propose to fully decouple person search towards optimal person search. A task-incremental person search network is proposed to incrementally construct an end-to-end model for the detection and re-id sub-task, which decouples the model architecture for the two sub-tasks. The proposed task-incremental network allows task-incremental training for the two conflicting tasks. This enables independent learning for different objectives thus fully decoupled the model for persons earch. Comprehensive experimental evaluations demonstrate the effectiveness of the proposed fully decoupled models for end-to-end person search., Comment: DICTA 2023 Best Student Paper

The widespread adoption of DNNs in NLP software has highlighted the need for robustness. Researchers proposed various automatic testing techniques for adversarial test cases. However, existing methods suffer from two limitations: weak error-discovering capabilities, with success rates ranging from 0% to 24.6% for BERT-based NLP software, and time inefficiency, taking 177.8s to 205.28s per test case, making them challenging for time-constrained scenarios. To address these issues, this paper proposes LEAP, an automated test method that uses LEvy flight-based Adaptive Particle swarm optimization integrated with textual features to generate adversarial test cases. Specifically, we adopt Levy flight for population initialization to increase the diversity of generated test cases. We also design an inertial weight adaptive update operator to improve the efficiency of LEAP's global optimization of high-dimensional text examples and a mutation operator based on the greedy strategy to reduce the search time. We conducted a series of experiments to validate LEAP's ability to test NLP software and found that the average success rate of LEAP in generating adversarial test cases is 79.1%, which is 6.1% higher than the next best approach (PSOattack). While ensuring high success rates, LEAP significantly reduces time overhead by up to 147.6s compared to other heuristic-based methods. Additionally, the experimental results demonstrate that LEAP can generate more transferable test cases and significantly enhance the robustness of DNN-based systems., Comment: Accepted at ASE 2023

Deep neural networks (DNNs) have been widely and successfully adopted and deployed in various applications of speech recognition. Recently, a few works revealed that these models are vulnerable to backdoor attacks, where the adversaries can implant malicious prediction behaviors into victim models by poisoning their training process. In this paper, we revisit poison-only backdoor attacks against speech recognition. We reveal that existing methods are not stealthy since their trigger patterns are perceptible to humans or machine detection. This limitation is mostly because their trigger patterns are simple noises or separable and distinctive clips. Motivated by these findings, we propose to exploit elements of sound ($e.g.$, pitch and timbre) to design more stealthy yet effective poison-only backdoor attacks. Specifically, we insert a short-duration high-pitched signal as the trigger and increase the pitch of remaining audio clips to `mask' it for designing stealthy pitch-based triggers. We manipulate timbre features of victim audios to design the stealthy timbre-based attack and design a voiceprint selection module to facilitate the multi-backdoor attack. Our attacks can generate more `natural' poisoned samples and therefore are more stealthy. Extensive experiments are conducted on benchmark datasets, which verify the effectiveness of our attacks under different settings ($e.g.$, all-to-one, all-to-all, clean-label, physical, and multi-backdoor settings) and their stealthiness. The code for reproducing main experiments are available at \url{https://github.com/HanboCai/BadSpeech_SoE}., Comment: 13 pages

Arising from the extreme/saddle point in electronic bands, Van Hove singularity (VHS) manifests divergent density of states (DOS) and induces various new states of matter such as unconventional superconductivity. VHS is believed to exist in one and two dimensions, but rarely found in three dimension (3D). Here, we report the discovery of 3D VHS in a topological magnet EuCd2As2 by magneto-infrared spectroscopy. External magnetic fields effectively control the exchange interaction in EuCd2As2, and shift 3D Weyl bands continuously, leading to the modification of Fermi velocity and energy dispersion. Above the critical field, the 3D VHS forms and is evidenced by the abrupt emergence of inter-band transitions, which can be quantitatively described by the minimal model of Weyl semimetals. Three additional optical transitions are further predicted theoretically and verified in magneto-near-infrared spectra. Our results pave the way to exploring VHS in 3D systems and uncovering the coordination between electronic correlation and the topological phase.

Keyword spotting (KWS) based on deep neural networks (DNNs) has achieved massive success in voice control scenarios. However, training of such DNN-based KWS systems often requires significant data and hardware resources. Manufacturers often entrust this process to a third-party platform. This makes the training process uncontrollable, where attackers can implant backdoors in the model by manipulating third-party training data. An effective backdoor attack can force the model to make specified judgments under certain conditions, i.e., triggers. In this paper, we design a backdoor attack scheme based on Voiceprint Selection and Voice Conversion, abbreviated as VSVC. Experimental results demonstrated that VSVC is feasible to achieve an average attack success rate close to 97% in four victim models when poisoning less than 1% of the training data., Comment: 7 pages,5 figures

Claim frequency data in insurance records the number of claims on insurance policies during a finite period of time. Given that insurance companies operate with multiple lines of insurance business where the claim frequencies on different lines of business are often correlated, multivariate count modeling with dependence for claim frequency is therefore essential. Due in part to the operation of bonus-malus systems, claims data in automobile insurance are often characterized by an excess of common zeros. This feature is referred to as multivariate zero-inflation. In this paper, we establish two ways of dealing with this feature. The first is to use a multivariate zero-inflated model, where we artificially augment the probability of common zeros based on standard multivariate count distributions. The other is to apply a multivariate zero-modified model, which deals with the common zeros and the number of claims incurred in each line, given that at least one claim occurs separately. A comprehensive comparative analysis of several models under these two frameworks is conducted using the data of an automobile insurance portfolio from a major insurance company in Spain. A less common situation in insurance is the absence of some common zeros resulting from incomplete records. This feature of these data is known as multivariate zero-deflation. In this case, our proposed multivariate zero-modified model still works, as shown by the second empirical study.

Keyword spotting (KWS) has been widely used in various speech control scenarios. The training of KWS is usually based on deep neural networks and requires a large amount of data. Manufacturers often use third-party data to train KWS. However, deep neural networks are not sufficiently interpretable to manufacturers, and attackers can manipulate third-party training data to plant backdoors during the model training. An effective backdoor attack can force the model to make specified judgments under certain conditions, i.e., triggers. In this paper, we design a backdoor attack scheme based on Pitch Boosting and Sound Masking for KWS, called PBSM. Experimental results demonstrated that PBSM is feasible to achieve an average attack success rate close to 90% in three victim models when poisoning less than 1% of the training data., Comment: 5 pages, 4 figures

The quest for realizing and manipulating ever smaller man-made movable structures and dynamical machines has spurred tremendous endeavors, led to important discoveries, and inspired researchers to venture to new grounds. Scientific feats and technological milestones of miniaturization of mechanical structures have been widely accomplished by advances in machining and sculpturing ever shrinking features out of bulk materials such as silicon. With the flourishing multidisciplinary field of low-dimensional nanomaterials, including one-dimensional (1D) nanowires/nanotubes, and two-dimensional (2D) atomic layers such as graphene/phosphorene, growing interests and sustained efforts have been devoted to creating mechanical devices toward the ultimate limit of miniaturization--genuinely down to the molecular or even atomic scale. These ultrasmall movable structures, particularly nanomechanical resonators that exploit the vibratory motion in these 1D and 2D nano-to-atomic-scale structures, offer exceptional device-level attributes, such as ultralow mass, ultrawide frequency tuning range, broad dynamic range, and ultralow power consumption, thus holding strong promises for both fundamental studies and engineering applications. In this Review, we offer a comprehensive overview and summary of this vibrant field, present the state-of-the-art devices and evaluate their specifications and performance, outline important achievements, and postulate future directions for studying these miniscule yet intriguing molecular-scale machines., Comment: 41 pages, 30 figures