arXiv subjects

Liran Ma

Publications and source records attributed to Liran Ma.

9 recordsLinked to original sources

arXiv2026

Masking Is Not Enough: Generative Restoration for Multimodal De-Identification in Medical AI

Medical image-text data can expose protected health information (PHI) through both visible image content as well as accompanying text, creating a barrier to privacy-preserving medical AI systems. This risk is especially prominent in multimodal systems, where images, questions, reports, and clinical context may enter training, evaluation, or inference pipelines. Existing medical vision-language benchmarks primarily emphasize task utility, while de-identification methods are often evaluated separately from downstream reasoning. We introduce ClinX, an end-to-end multimodal PHI sanitization framework for medical image-text data. ClinX detects visible identifiers with optical character recognition (OCR), constructs binary PHI masks, and applies ClinX-PRISM, a no-skip generative restoration module with privacy-oriented post-processing for burned-in identifier suppression. In parallel, text-side PHI is reduced through progressive de-identification levels: regex masking, context-aware masking, and rewrite-based sanitization. We evaluate ClinX in medical visual question answering (MedVQA), jointly measuring PHI leakage and downstream utility across image-side, text-side, and combined de-identification settings. Results show that OCR-only masking is not sufficient as a standalone solution, and restoration-based sanitization better preserves clinically relevant visual context while sharply reducing recoverable PHI.

cs.CV↗

arXiv2026

Q-Interference: Memory-Efficient Phase-Aware Quantum-Inspired Attention

GPT attention measures token compatibility through dot-product similarity. This mechanism is simple, effective, and memory-efficient. But it does not explicitly model whether strong token features should reinforce or suppress one another. We introduce Q-Interference, a fully classical quantum-inspired attention mechanism for autoregressive language modeling that augments each query and key feature with an amplitude and a learned phase. The resulting attention score is phase-aware which aligned phases contribute constructively while conflicting phases contribute destructively. Although Q-Interference yields a richer interaction rule than similarity alone, a naive implementation of Q-Interference requires a large token-pair-feature interaction tensor, making it memory-intensive and often impractical. To address this limitation, we propose an exact trigonometric factorization that computes the same score using two standard matrix multiplications avoiding materialization of the large intermediate tensor. Q-Interference fits directly into a Transformer block in GPT and leaves the remainder of the model architecture and next-token prediction objective unchanged. Experiments on public benchmark datasets and baseline models show that the proposed reformulation trains stably in a controlled GPT-style setting and provides a consistent memory advantage over naive phase-aware interference attention. These results support the specific contribution of this work: an exact memory-efficient reformulation that makes phase-aware interference attention practical within a standard GPT pipeline.

cs.CL↗

arXiv2026

Reliability-Aware LLM Alignment from Inconsistent Human Feedback

Reinforcement Learning from Human Feedback (RLHF) is critical for aligning Large Language Models (LLMs) with human preferences. However, its efficacy is often compromised by the inherent inconsistency and subjectivity of human annotations. Existing preference optimization frameworks, such as Direct Preference Optimization (DPO), typically treat ambiguous pairs with high annotator disagreement identically to those with unanimous consensus, forcing models to overfit to inconsistent supervision signals and leading to suboptimal alignment. In this work, we propose Reliability-Guided Preference Optimization (RGPO), a robust framework designed to mitigate the impact of inconsistent human feedback. RGPO estimates annotator reliability and infers latent ground truth labels from noisy human feedback to identify robust preferences. Furthermore, we introduce a reliability-aware consistency optimization that dynamically modulates the training objective based on the consensus level of annotations, ensuring the model prioritizes high-consensus supervision signals. Extensive experiments on LLM alignment benchmarks demonstrate that RGPO effectively reduces inconsistency and noise in training data and achieves superior performance compared to widely adopted RLHF baselines. Our code and configurations are available at https://github.com/GenieHuang/RGPO.

cs.AI↗

arXiv2025

Variations of saturation vapor pressure and evaporation rate of liquids with their vaporization enthalpy

The phase transition enthalpy of condensed materials can be altered by factors such as electric fields, and such variations in turn affect physical and chemical behaviors including phase equilibrium. However, due to the neglect of non-expansion work, the Clapeyron equation does not account for the effect of changes in phase transition enthalpy on equilibrium. In this paper, by analyzing the electric-field-induced changes in phase transition enthalpy and incorporating the non-expansion work performed on the system, we extended both the Clapeyron and Clausius-Clapeyron equations to explicitly include variations in phase transition enthalpy. Building upon these extensions, analytical expressions for the vapor pressure of liquids and for the total evaporation rate of sessile liquid droplets as functions of the change in vaporization enthalpy have been derived, showing that an approximately 1% decrease in vaporization enthalpy of a liquid can cause an increase of about 20% in its vapor pressure and an almost twofold increase in its evaporation rate. The theoretical predictions by the present equations were compared and found to be in good agreement with the experimental observations.

cond-mat.soft↗

arXiv2024

Effect of UV light irradiation on charge neutralization in XPS measurements

When XPS analyses are performed on insulator surfaces, shift and deformation of spectra peaks typically take place due to the surface charging. To achieve reliable XPS measurements, neutralization techniques have been widely adopted but their effectiveness are still limited, and thus, new neutralization technologies are urgently needed. Here, stable XPS spectra in which all the peaks undergo a reduced and nearly constant shift without significant deformation and broadening were obtained by introducing the UV light irradiation, implying that the introduction of the UV light can not only greatly attenuate the strength but also significantly improve both the temporal stability and the spatial uniformity of the surface charging during XPS measurements. This phenomenon, referred to as UV-assisted neutralization in this article, was found as effective as the most commonly used dual beam charge neutralization. Further observations show that the suppression of the charging issue comes from the adsorption of the UV-excited photoelectrons onto the X-ray irradiation region. This neutralization method, combined with the binding energy referencing, can be expected to become a promising alternative technique for solving the charging issues in XPS measurements.

physics.ins-det↗

arXiv2021

Cross-layer based intermittent jamming schemes for securing energy-constraint networks

The Internet-of-Things (IoT) emerges as a paradigm to achieve ubiquitous connectivity via wireless communications between kinds of physical objects. Due to the wireless broadcasting nature and the energy constraint of physical objects, concerns on IoT security have triggered research on cooperative jamming based physical layer security. With the help of a cooperative jammer, existing solutions can effectively fight against eavesdroppers. However, these schemes are of high energy cost due to continuously transmitting jamming signals. To reduce the energy consumption, we propose a new idea of intermittent jamming and design five specific intermittent jamming schemes (IJSs). By taking the transmit frame formate into account, we optimize these IJSs from three aspects, including the jamming power, the jamming method, and the jamming positions. Then we analyze the applicability of the proposed IJSs according to different requirements on the synchronization, the available jamming energy and the jamming power constraints. Extensive MATLAB experiments are conducted on the basis of the WLAN Toolbox, which demonstrate the proposed IJSs can effectively degrade the reception of the eavesdropper and outperform the widespread continuous jamming scheme (CJS) when the available jamming energy is limited.

cs.IT↗

arXiv2020

An easily prepared, simple information storage and display device based on triboelectric enhanced mechanoluminescence phenomenon and droplet-luminescence

One easy-make, flexible information storage and display device is prepared. The device show memory effect, which can record the friction trace in real time through the triboelectric effect, and then reproduce the trajectory under the excitation of any kind of polar liquid droplet, completely display the triboelectric area. The explanation mechanism of this phenomenon is proposed, and the main influencing factors of the frictional electrification between solid materials and flexible substrates are determined by systematic experiements. This phenomenon connects mechanoluminescence, electroluminescence, and triboelectrification. It is of great significance to the research of triboelectricity and mechanoluminescence, and is expected to play an important role in display, information storage and encryption.

cond-mat.mtrl-sci↗

arXiv2019

The electron transfer dynamics in the contact electrification and its effects on the intensity of triboluminescence

With the growing threat of energy crisis and the increasing need to power microelectronic devices, people are seeking potential alternative energies that can replace the conventional sources such as fossil fuels. Due to its simple structure, low cost, and high performance, the triboelectric nanogenerator (TENG) which is based on the contact electrification has become one of the most promising candidates. Although plays a crucial role in determining the amount of charge transfer of TENG, the fundamental mechanism that underlies the charge transfer in the contact electrification of insulator-insulator contacts has still not been completely understood. In this paper, the dynamics of charge transfer in the contact electrification is investigated by observing the triboluminescence of the sliding contacts and a theoretical model is proposed for the electrification of insulator-insulator contacts. In the present "Capacitor model", the electron tunneling is assumed to be the dominating mechanism in the electrification and the driving force for the electron transfer is the contact potential difference, and thus, the electrification is considered as a charging process of the capacitor formed by the contact surfaces. The consistency between the predictions by the present model and the experiments may confirm again that the electron transfer is the dominant process in the contact electrification of insulator-insulator contacts.

physics.app-ph↗

arXiv2017

Effects of the Position Reversal of Friction Pairs on the Strength of Tribocharging and Tribodischarging

The friction-induced charging (i.e., tribocharging) and the following discharging (referred here as tribodischarging) are always believed to have negative effects on the daily life and on the industrial production. Thus, how to inhibit the tribocharging and the tribodischarging has caused wide public concern. Because the discharge caused by the electrical breakdown of the ambient gas is generally accompanied with the generation of light, we investigated here the tribocharging and the tribodischarging by observing the light emitted during friction. We found that the position reversal of the friction pair has a dramatic impact on the intensity of the tribo-induced light. Experimental results show that an intense light is produced when a stationary Al2O3 disk is sliding on a rotating SiO2 disk, but only a weak light is observed for the case of a stationary SiO2 disk and a rotating Al2O3 disk. This means that the process of the tribocharging and the tribodischarging can be significantly influenced owing to the change in the relative position of the friction couple. The experimentally measured polarities of the tribo-induced charge on the friction surfaces further indicated that the strong discharging occurs when the rotating surface is negatively charged. The reason for the difference in the intensity of the tribocharging and tribodischarging can be attributed to the combined effects of the contact potential difference and the temperature gradient between the contacting surfaces on the charge transfer when friction. Finally, a simple, low cost, yet effective approach, i.e., just keep the friction partner whose surface is tribo-induced negatively charged as the stationary one, can be utilized to suppress the intensity of the tribocharging and the tribodischarging. This work may provide potential applications in numerous areas of science and engineering and also in the everyday life.

cond-mat.mtrl-sci↗