arXiv subjects

Xinlei Liu

Publications and source records attributed to Xinlei Liu.

9 recordsLinked to original sources

arXiv2026

MAPE: Defending Against Transferable Adversarial Attacks Using Multi-Source Adversarial Perturbations Elimination

Neural networks are vulnerable to meticulously crafted adversarial examples, leading to high-confidence misclassifications in image classification tasks. Due to their consistency with regular input patterns and the absence of reliance on the target model and its output information, transferable adversarial attacks exhibit a notably high stealthiness and detection difficulty, making them a significant focus of defense. In this work, we propose a deep learning defense known as multi-source adversarial perturbations elimination (MAPE) to counter diverse transferable attacks. MAPE comprises the single-source adversarial perturbation elimination (SAPE) mechanism and the pre-trained models probabilistic scheduling algorithm (PPSA). SAPE utilizes a thoughtfully designed channel-attention U-Net as the defense model and employs adversarial examples generated by a pre-trained model (e.g., ResNet) for its training, thereby enabling the elimination of known adversarial perturbations. PPSA introduces model difference quantification and negative momentum to strategically schedule multiple pre-trained models, thereby maximizing the differences among adversarial examples during the defense model's training and enhancing its robustness in eliminating adversarial perturbations. MAPE effectively eliminates adversarial perturbations in various adversarial examples, providing a robust defense against attacks from different substitute models. In a black-box attack scenario utilizing ResNet-34 as the target model, our approach achieves average defense rates of over 95.1\% on CIFAR-10 and over 71.5\% on Mini-ImageNet, demonstrating state-of-the-art performance.

cs.CV↗

arXiv2026

SDM: A Powerful Tool for Evaluating Model Robustness

Gradient-based attacks are important methods for evaluating model robustness. However, since the proposal of APGD, it has been difficult for such methods to achieve significant breakthroughs. To achieve such an effect, we first analyze the issue of "high-loss non-adversarial examples" that degrades attack performance in previous methods, and prove that this issue arises from inappropriate objectives for adversarial example generation. Subsequently, we reconstruct the objective as "maximizing the difference between the non-ground-truth label probability upper bound and the ground-truth label probability", and proposes a novel and powerful gradient-based attack method named Sequential Difference Maximization (SDM). SDM establishes a three-layer optimization framework of "cycle-stage-step". It adopts the negative probability loss function and the Directional Probability Difference Ratio (DPDR) loss function in the initial and subsequent optimization stages, respectively, and approaches the ideal objective of adversarial example generation via stage-wise sequential optimization. Experiments demonstrate that compared with previous state-of-the-art methods, SDM not only achieves stronger attack performance but also exhibits superior cost-effectiveness. The code is available at https://github.com/X-L-Liu/ICML-SDM.

cs.CV↗

arXiv2026

Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learning

Reinforcement learning in massively parallel physics simulations has driven major progress in sim-to-real robot learning. However, current approaches remain brittle and task-specific, relying on extensive per-task engineering to design rewards, curricula, and demonstrations. Even with this engineering, they often fail on long-horizon, contact-rich manipulation tasks and do not meaningfully scale with compute, as performance quickly saturates when training revisits the same narrow regions of state space. We introduce OmniReset, a simple and scalable framework that enables on-policy reinforcement learning to robustly solve a broad class of dexterous manipulation tasks using a single reward function, fixed algorithm hyperparameters, no curricula, and no human demonstrations. Our key insight is that long-horizon exploration can be dramatically simplified by using simulator resets to systematically expose the RL algorithm to the diverse set of robot-object interactions which underlie dexterous manipulation. OmniReset programmatically generates such resets with minimal human input, converting additional compute directly into broader behavioral coverage and continued performance gains. We show that OmniReset gracefully scales to long-horizon dexterous manipulation tasks beyond the capabilities of existing approaches and is able to learn robust policies over significantly wider ranges of initial conditions than baselines. Finally, we distill OmniReset into visuomotor policies which display robust retrying behavior and substantially higher success rates than baselines when transferred to the real world zero-shot. Project webpage: https://weirdlabuw.github.io/omnireset/

cs.RO↗

arXiv2026

Symbol-Synchronous Communication for Ultra-Low-Power Multi-Hop Ambient IoT Networks

Ambient Internet of Things (A-IoT) devices, as a critical enabler of future green IoT networks, have attracted broad interest from both industry and academia due to their ability to operate without batteries and with low maintenance costs. To accommodate their dynamic and constrained energy budget, an ultra-low-power connectivity protocol is required. Due to the severely limited transmit power of A-IoT devices, multi-hop connectivity is an interesting paradigm to extend their range. However, commonly used protocols for multi-hop communication may not be suitable for A-IoT due to excessive overhead related to channel access procedures, coordinated routing, and tight time synchronization requirements. This paper presents a novel network connectivity protocol based on symbol-synchronous transmissions, which allows battery-less relay nodes to participate in the forwarding process in an ad-hoc manner, without the need for synchronization or coordination. This allows them to adapt their duty cycle to the available harvested energy. Simulation results show that the proposed protocol can ensure high reliability in data packet delivery while significantly reducing the energy consumption of each relay node. We also investigate the relationship between wake-up probability and network density. For example, a 400-node network in a 625 m2 area can achieve a packet error rate below 1 % with an average awake time of 6 % per node, achieving an energy consumption reduction of 88 % compared to the baseline approach.

cs.NI↗

arXiv2026

RF-Zero-Wire: Design and Analysis of Multi-Hop Low-latency Symbol-synchronous RF Communication

The latency gap between wired and wireless networks poses a challenge in the adoption of wireless technologies in latency-sensitive scenarios. The gap is especially notable in multi-hop communication typical for industrial sensor networks and robotic swarms. The main reason behind it is that commonly used wireless protocols rely on store-and-forward routing and costly overhead procedures to avoid interference. This article introduces RF-Zero-Wire, an RF-based symbol-synchronous communication protocol. Instead of relaying the whole frame per hop in a store-and-forward manner, nodes concurrently relay the frame symbol by symbol, without the need for tight time synchronization. Based on data collected in real-world experiments, we reveal that the inevitable carrier frequency offsets (CFOs) introduced by imperfect crystal oscillators cause a beating effect under concurrent symbol transmissions. This is characterized by periodic constructive and destructive interference, which significantly affects reliability. Subsequently, a thorough simulation study shows how the beating problem can be overcome with error correction codes. RF-Zero-Wire allows achieving an end-to-end latency of less than 1ms for a small 4-byte frame transmitted across 5 hops. Moreover, latency is shown to increase only by 0.16% per extra hop for 16-byte frames, which is negligible compared to the over 100% per-hop latency increase observed in store-and-forward protocols. The trade-offs between network reliability and CFO range, communication distance, node density, and achievable data rate are studied in large-scale experiments based on simulation.

cs.NI↗

arXiv2026

Advances in non-Hermitian dynamics of quadratic bosonic systems

Non-Hermitian physics has emerged as a rapidly advancing field of research, revealing a range of novel phenomena and potential applications. Traditional non-Hermitian Hamiltonians are typically simulated by constructing asymmetric couplings or by introducing dissipation and gain to realize non-Hermitian systems. The quadratic bosonic system (QBS) with squeezing interaction is intrinsically Hermitian; however, its dynamical evolution matrix in both real and momentum spaces is non-Hermitian. Based on this, applying a field-operator transformation xp to the dynamical evolution matrix yields quadrature nonreciprocal transmission between the x and p operators. This nonreciprocal characteristic can be utilized in signal amplifiers. On the other hand, within the Bogoliubov-de Gennes framework in momentum space, one can observe non-Hermitian topological phenomena such as point-gap topology and the non-Hermitian skin effect, both induced by spectra with nonzero winding numbers. Additionally, QBS can be employed to realize non-Hermitian Aharonov-Bohm cages and to extend non-Bloch band theory. Previous studies in non-Hermitian physics have largely concentrated on classical systems. The influence of non-Hermitian properties on quantum effects remains a key issue awaiting exploration and has evolved into a research direction at the interface of non-Hermitian and quantum physics.

quant-ph↗

arXiv2025

Sequential Difference Maximization: Generating Adversarial Examples via Multi-Stage Optimization

Efficient adversarial attack methods are critical for assessing the robustness of computer vision models. In this paper, we reconstruct the optimization objective for generating adversarial examples as "maximizing the difference between the non-true labels' probability upper bound and the true label's probability," and propose a gradient-based attack method termed Sequential Difference Maximization (SDM). SDM establishes a three-layer optimization framework of "cycle-stage-step." The processes between cycles and between iterative steps are respectively identical, while optimization stages differ in terms of loss functions: in the initial stage, the negative probability of the true label is used as the loss function to compress the solution space; in subsequent stages, we introduce the Directional Probability Difference Ratio (DPDR) loss function to gradually increase the non-true labels' probability upper bound by compressing the irrelevant labels' probabilities. Experiments demonstrate that compared with previous SOTA methods, SDM not only exhibits stronger attack performance but also achieves higher attack cost-effectiveness. Additionally, SDM can be combined with adversarial training methods to enhance their defensive effects. The code is available at https://github.com/X-L-Liu/SDM.

cs.CV↗

arXiv2024

Low-latency Symbol-Synchronous Communication for Multi-hop Sensor Networks

Wireless sensor networks (WSNs) have received great interest due to their scalability, energy efficiency, and low-cost deployment. By utilizing multi-hop communication, WSNs can cover a wide area using low transmission power without the need for any communication infrastructure. Traditionally, WSNs rely on store-and-forward routing protocols and Time Division Multiple Access (TDMA)-based schedules that avoid interference between different wireless nodes. However, emerging challenging scenarios, such as the industrial Internet of Things (IoT) and robotic swarms, impose strict latency and reliability requirements, which traditional approaches cannot fulfill. In this paper, we propose a novel symbol-synchronous transmission design that provides reliable low-latency communication with a reasonable data rate on classical sub-6GHz RF frequency bands (e.g., the 2.4GHz ISM band). Instead of avoiding overlapping transmissions, the proposed scheme benefits from concurrent transmissions. Using simulation in MATLAB, we prove that the proposed design allows achieving a wire-like delay of 5ms for a 512-bit packet over multiple hops with only a 0.3% latency increase per extra hop and a low bit error rate (BER) of 0.04%. Compared to similar state-of-the-art approaches it can achieve a significantly higher data rate of 100kbps, which is expected to increase further with future improvements of the system.

cs.NI↗

arXiv2021

On the Formation of Hydrogen Peroxide in Water Microdroplets

Recent reports on the formation of hydrogen peroxide (H$_2$O$_2$) in water microdroplets produced via pneumatic spraying or capillary condensation have garnered significant attention. How covalent bonds in water could break under such conditions challenges our textbook understanding of physical chemistry and the water substance. While there is no definitive answer, it has been speculated that ultrahigh electric fields at the air-water interface are responsible for this chemical transformation. Here, we resolve this mystery via a comprehensive experimental investigation of H$_2$O$_2$ formation in (i) water microdroplets sprayed over a range of liquid flow-rates, the (shearing) air flow rates, and the air composition (ii) water microdroplets condensed on hydrophobic substrates formed via hot water or humidifier under controlled air composition. Specifically, we assessed the contributions of the evaporative concentration and shock waves in sprays and the effects of trace O$_3$(g) on the H$_2$O$_2$ formation. Glovebox experiments revealed that the H$_2$O$_2$ formation in water microdroplets was most sensitive to the air-borne ozone (O$_3$) concentration. In the absence of O$_3$(g), we could not detect H$_2$O$_2$(aq) in sprays or condensates (detection limit: 250 nM. In contrast, microdroplets exposed to atmospherically relevant O$_3$(g) concentration (10-100 ppb) formed 2-30 $μ$M H$_2$O$_2$(aq); increasing the gas-liquid surface area, mixing, and contact duration increased H$_2$O$_2$(aq) concentration. Thus, the mystery is resolved - the water surface facilitates the O$_3$(g) mass transfer, which is followed by the chemical transformation of O$_3$(aq) into H$_2$O$_2$(aq). These findings should also help us understand the implications of this chemistry in natural and applied contexts.

cond-mat.soft↗