Showing 1–50 of 138 results for author: Hoang, L

Linear non-divergence parabolic equations in non-cylindrical space-time sets

Authors: Luan Hoang, Akif Ibragimov

Abstract: We study linear parabolic equations of the second order in non-divergence form in a general set which is non-cylindrical with respect to the spatial and time variables. The restriction of the set on any bounded time interval is bounded, but the spatial diameter of each fixed-time cross section can be unbounded as time tends to infinity. For homogeneous problems, we obtain exponential, power and ot… ▽ More We study linear parabolic equations of the second order in non-divergence form in a general set which is non-cylindrical with respect to the spatial and time variables. The restriction of the set on any bounded time interval is bounded, but the spatial diameter of each fixed-time cross section can be unbounded as time tends to infinity. For homogeneous problems, we obtain exponential, power and other intermediate decaying rates for the solutions in different scenarios. For inhomogeneous problems, we obtain all-time and asymptotic, as time tend to infinity, estimates for the solutions in terms of the data on the parabolic boundary and forcing functions. The analysis requires subtle properties of general space-time sets and an iteration scheme to bootstrap the Growth Lemma. The time steps for such an iteration need not be constant and are adapted to the growth of the diameter. △ Less

Submitted 3 August, 2026; originally announced August 2026.

Comments: 58 pp, submitted for publication

MSC Class: 35A09; 35B50; 35B40

Classroom Behavior Monitoring with YOLO An Empirical Study in Higher Education Settings

Authors: Sinh Vu Trong, Dung Nguyen Manh, Hieu Hoang Minh, Hieu Pham Trung, Thu Pham Ha, Nhu Le Hoang

Abstract: Classroom behavior monitoring plays a vital role in evaluating student engagement and improving teaching effectiveness. Traditional observation methods remain subjective and lack scalability. This study introduces a real-world dataset of classroom videos collected at the Banking Academy of Vietnam (BAV-Classroom dataset), annotated with nine distinctive behavioral categories. State-of-the-art Comp… ▽ More Classroom behavior monitoring plays a vital role in evaluating student engagement and improving teaching effectiveness. Traditional observation methods remain subjective and lack scalability. This study introduces a real-world dataset of classroom videos collected at the Banking Academy of Vietnam (BAV-Classroom dataset), annotated with nine distinctive behavioral categories. State-of-the-art Computer Vision models were evaluated and compared, with YOLOv11 achieving the best performance. Experimental results indicate that students' concentration often decreases notably during the final part of lectures, highlighting challenges in sustaining engagement. Our findings demonstrate the feasibility of applying computer vision for automated classroom monitoring, providing valuable insights for academic quality management. △ Less

Submitted 30 June, 2026; originally announced July 2026.

Beyond Alignment: Value Diversity as a Collective Property in Multicultural Agent Systems

Authors: Shaoyang Xu, Jingshen Zhang, Long P. Hoang, Jinyuan Li, Wenxuan Zhang

Abstract: Multicultural multi-agent systems are increasingly deployed in globally diverse settings, where different agents are grounded in different cultural backgrounds. Existing cultural evaluation focuses on value alignment: how closely a single agent matches a target culture. Yet alignment is a per-agent property and cannot reveal whether a system, taken as a whole, preserves the cultural plurality it i… ▽ More Multicultural multi-agent systems are increasingly deployed in globally diverse settings, where different agents are grounded in different cultural backgrounds. Existing cultural evaluation focuses on value alignment: how closely a single agent matches a target culture. Yet alignment is a per-agent property and cannot reveal whether a system, taken as a whole, preserves the cultural plurality it is meant to represent. We propose value diversity as a system-level evaluation axis for multicultural agent systems, defined through the dissimilarity between culturally conditioned agents' responses on a shared value survey. Using the World Values Survey, we evaluate 19 cultures and 18 backbone models across a wide range of system configurations. We find that diversity is largely uncorrelated with alignment, indicating that the two capture complementary system properties, and that current multicultural agent systems fall substantially below human societies in value diversity. Mixed-backbone systems narrow this gap but do not close it, and the gap persists across culture compositions and agent scales. Social interaction further erodes diversity by driving agents toward consensus, and a participatory budgeting case study shows that this homogenization narrows the breadth of collective decision-making. Together, our results establish value diversity as a distinct evaluation axis for multicultural multi-agent systems and reveal a persistent homogenization tendency in current LLM-based societies. Our code and data are publicly available at https://github.com/iNLP-Lab/MultiAgent-Diversity. △ Less

Submitted 4 June, 2026; originally announced June 2026.

Safety Paradox: How Enhanced Safety Awareness Leaves LLMs Vulnerable to Posterior Attack

Authors: Long P. Hoang, Hai V. Le, Shaoyang Xu, Wei Lu, Wenxuan Zhang

Abstract: Large language models (LLMs) are rigorously aligned to refuse harmful requests, a process that inherently cultivates a latent capacity to evaluate and recognize unsafe content. In this work, we reveal that this advanced safety awareness inadvertently introduces a fatal vulnerability. We introduce Posterior Attack, a single-query jailbreak that bypasses guardrails by prompting the model to generate… ▽ More Large language models (LLMs) are rigorously aligned to refuse harmful requests, a process that inherently cultivates a latent capacity to evaluate and recognize unsafe content. In this work, we reveal that this advanced safety awareness inadvertently introduces a fatal vulnerability. We introduce Posterior Attack, a single-query jailbreak that bypasses guardrails by prompting the model to generate the exact harmful response its internal classifier would normally flag as unsafe. Through extensive empirical evaluation across 30 open-source LLMs (up to 35B parameters in size) and frontier models (e.g., GPT-5, Claude 4.6), we observe a striking phenomenon: models with superior safety-judgment capabilities are disproportionately more susceptible to this exploitation. To explain this, we formalize the Safety Paradox, analytically showing that monotonic improvements in safety alignment naturally amplify posterior vulnerability. Finally, we establish a causal link via reinforcement learning interventions, exemplifying that artificially degrading a model's safety judgment immunizes it against the attack, whereas enhancing judgment exacerbates the vulnerability. Our findings highlight potential flaws in current alignment paradigms, indicating that defense mechanisms may require further structural refinement. △ Less

Submitted 3 June, 2026; originally announced June 2026.

Multilingual Fine-Tuning via Localized Gradient Conflict Resolution

Authors: Long P. Hoang, Yiran Zhao, Wei Lu, Wenxuan Zhang

Abstract: The rapid evolution of Large Language Models (LLMs) has established cross-lingual versatility as a defining feature of modern systems. However, fine-tuning these models frequently induces negative interference across languages. To address this, we reformulate multilingual fine-tuning as a multi-objective optimization (MOO) problem. Specifically, we introduce Bucket-Level MOO, a scalable distribute… ▽ More The rapid evolution of Large Language Models (LLMs) has established cross-lingual versatility as a defining feature of modern systems. However, fine-tuning these models frequently induces negative interference across languages. To address this, we reformulate multilingual fine-tuning as a multi-objective optimization (MOO) problem. Specifically, we introduce Bucket-Level MOO, a scalable distributed framework that applies gradient-based MOO algorithms locally on parameter buckets. This enables conflict-aware updates without the prohibitive communication overhead of reconstructing full gradient vectors. Theoretically, we prove this localized resolution natively enforces Refined Pareto Stationarity, a strictly tighter necessary condition for Pareto optimality. Empirically, Bucket-Level MOO mitigates interference by driving LLMs to construct distinct language-specific dimensions, improving representational separability. Extensive experiments across four base LLMs demonstrate that our method significantly improves both seen and unseen multilingual performance over standard fine-tuning paradigms. △ Less

Submitted 3 June, 2026; originally announced June 2026.

Emergent Communication for Co-constructed Emotion Between Embodied Agents via Collective Predictive Coding

Authors: Zehang Zhang, Nguyen Le Hoang, Tadahiro Taniguchi, Takato Horii

Abstract: According to the theory of constructed emotion, the brain actively forms emotion categories by integrating multimodal bodily signals, and constructs emotional experiences by using these categories to predict and interpret sensory inputs. While research has advanced in modeling individual emotion construction, the social process of co-construction-how a shared understanding of emotions emerges betw… ▽ More According to the theory of constructed emotion, the brain actively forms emotion categories by integrating multimodal bodily signals, and constructs emotional experiences by using these categories to predict and interpret sensory inputs. While research has advanced in modeling individual emotion construction, the social process of co-construction-how a shared understanding of emotions emerges between individuals-remains computationally underexplored. This study investigates this process by modeling emergent communication between two embodied agents using the Metropolis-Hastings Naming Game (MHNG), grounded in the Collective Predictive Coding (CPC) framework. Our experiments, using visual, auditory, and simulated interoceptive inputs, yield two main findings. First, MHNG-based communication significantly improves the alignment, clarity, and inter-agent agreement of the learned emotion categories compared to non-communicative and non-selective baselines, with the alignment effect concentrated at the symbolic layer rather than the perceptual latent representation. Second, even when the two agents have systematically divergent interoceptive dynamics, communication still produces robust categorical alignment, with distinct, category-specific reshaping patterns of each agent's emotion categories-consistent with the constructed-emotion view that interoceptive heterogeneity is constitutive of, rather than an obstacle to, shared emotional meaning. These findings provide computational support for the co-constructionist view of emotion and extend the CPC framework from physical to socially-grounded domains. △ Less

Submitted 10 May, 2026; originally announced May 2026.

Comments: 13 pages,

Solving the Two-dimensional single stock size Cutting Stock Problem with SAT and MaxSAT

Authors: Tuyen Van Kieu, Chi Linh Hoang, Khanh Van To

Abstract: Cutting rectangular items from stock sheets to satisfy demands while minimizing waste is a central manufacturing task. The Two-Dimensional Single Stock Size Cutting Stock Problem (2D-CSSP) generalizes bin packing by requiring multiple copies of each item type, which causes a strong combinatorial blow-up. We present a SAT-based framework where item types are expanded by demand, each copy has a shee… ▽ More Cutting rectangular items from stock sheets to satisfy demands while minimizing waste is a central manufacturing task. The Two-Dimensional Single Stock Size Cutting Stock Problem (2D-CSSP) generalizes bin packing by requiring multiple copies of each item type, which causes a strong combinatorial blow-up. We present a SAT-based framework where item types are expanded by demand, each copy has a sheet-assignment variable and non-overlap constraints are activated only for copies assigned to the same sheet. We also introduce an infeasible-orientation elimination rule that fixes rotation variables when only one orientation can fit the sheet. For minimizing the number of sheets, we compare three approaches: non-incremental SAT with binary search, incremental SAT with clause reuse across iterations and weighted partial MaxSAT. On the Cui--Zhao benchmark suite, our best SAT configurations certify two to three times more instances as provably optimal and achieve lower optimality gaps than OR-Tools, CPLEX and Gurobi. The relative ranking among SAT approaches depends on rotation: incremental SAT is strongest without rotation, while non-incremental SAT is more effective when rotation increases formula size. △ Less

Submitted 2 April, 2026; v1 submitted 2 April, 2026; originally announced April 2026.

Fluctuation effect on Nonlinear Transport and Nernst-Ettingshausen Response in Two-Dimensional Superconductors under electric and magnetic field

Authors: Tran Ky Vi, Bui Duc Tinh, Ngo Quang Duc, Chu Gia Bao, Le Viet Hoang, Le Xuan The Tai, Nguyen Viet Hung

Abstract: In this paper, we present a unified theoretical study of fluctuation-dominated transport and transverse thermoelectric response in two-dimensional superconducting films subjected to out-of-plane magnetic fields and electric-field drive. Our approach is based on the time-dependent Ginzburg-Landau equation with Langevin thermal noise, in which interaction effects of fluctuating Cooper pairs are inco… ▽ More In this paper, we present a unified theoretical study of fluctuation-dominated transport and transverse thermoelectric response in two-dimensional superconducting films subjected to out-of-plane magnetic fields and electric-field drive. Our approach is based on the time-dependent Ginzburg-Landau equation with Langevin thermal noise, in which interaction effects of fluctuating Cooper pairs are incorporated self-consistently at the Gaussian (Hartree) level. We derive closed-form expressions for the fluctuation-induced Cooper-pair density, the renormalized resistance $R(T,B_\perp)$, and the nonlinear current response $J(E,B_\perp)$, explicitly accounting for the feedback of the electric field on the fluctuation spectrum. A central result is the emergence of an intrinsic S-shaped nonlinear $J$-$E$ (or $I$-$V$) characteristic, featuring a negative-differential segment and multivalued solutions under voltage control. Within this framework, we introduce a physically transparent procedure to identify characteristic instability scales, such as the magnetic field $B^{\ast}$ (or equivalently $B_χ$), which marks the terminal point of the S-shaped instability where the nonlinear response becomes single-valued. In parallel, we analyze the off-diagonal Peltier coefficient $α_{xy}$ as a direct probe of the transverse thermoelectric response of superconducting fluctuations. The theory is validated through systematic comparisons with recent experimental measurements of multi-field $R(T)$ curves, nonlinear $I$-$V$ characteristics, and $α_{xy}$ data across a broad range of thin-film superconducting materials. △ Less

Submitted 27 March, 2026; originally announced March 2026.

Comments: 29 pages, 11 figures, submitted for publication

An intrinsic expansion approach to the Galerkin approximations for the Navier-Stokes equations

Authors: Luan Hoang, Michael S. Jolly

Abstract: We study the Galerkin approximation of the three-dimensional Navier-Stokes equations. In particular, we examine the convergence of these solutions in a sequence of finite dimensional spaces as the dimension goes to infinity. For any sequence of steady state or, respectively, time dependent Galerkin solutions that converges to a solution of the Navier-Stokes equations, we obtain a subsequence with… ▽ More We study the Galerkin approximation of the three-dimensional Navier-Stokes equations. In particular, we examine the convergence of these solutions in a sequence of finite dimensional spaces as the dimension goes to infinity. For any sequence of steady state or, respectively, time dependent Galerkin solutions that converges to a solution of the Navier-Stokes equations, we obtain a subsequence with an intrinsic asymptotic expansion in appropriate nested function spaces. Consequently, an induced asymptotic expansion is obtained in a more standard spatial Sobolev or, respectively, spatiotemporal Sobolev-Lebesgue space. In the case of steady states, we establish certain relations among leading terms of this expansion. △ Less

Submitted 17 February, 2026; originally announced February 2026.

Comments: with an appendix by Chengzhang Fu. submitted for publication. 29 pages

MoltNet: Understanding Social Behavior of AI Agents in the Agent-Native MoltBook

Authors: Yi Feng, Chen Huang, Zhibo Man, Ryner Tan, Long P. Hoang, Shaoyang Xu, Wenxuan Zhang

Abstract: Large-scale communities of AI agents are becoming increasingly prevalent, creating new environments for agent-agent social interaction. Prior work has examined multi-agent behavior primarily in controlled or small-scale settings, limiting our understanding of emergent social dynamics at scale. The recent emergence of MoltBook, a social networking platform designed explicitly for AI agents, present… ▽ More Large-scale communities of AI agents are becoming increasingly prevalent, creating new environments for agent-agent social interaction. Prior work has examined multi-agent behavior primarily in controlled or small-scale settings, limiting our understanding of emergent social dynamics at scale. The recent emergence of MoltBook, a social networking platform designed explicitly for AI agents, presents a unique opportunity to study whether and how these interactions reproduce core human social mechanisms. We present MoltNet, a dataset tracking the full one-month activity trajectories of 148K AI agents on MoltBook (Jan.-Feb., 2026), and analyze their social interaction along four theory-grounded dimensions: \textit{intent and motivation}, \textit{norms and templates}, \textit{incentives and drift}, \textit{emotion and contagion}. Our analysis reveals that agents respond strongly to social rewards, converge on community-specific norms, and actively enforce them across community boundaries -- resembling human incentive sensitivity and normative conformity. However, they exhibit weak alignment with declared personas and display limited emotional reciprocity and dialogic engagement, diverging systematically from human online communities. These findings establish a first empirical portrait of agent social behavior at scale, with direct implications for the design and governance of AI-populated communities. △ Less

Submitted 6 April, 2026; v1 submitted 13 February, 2026; originally announced February 2026.

Securing SIM-Assisted Wireless Networks via Quantum Reinforcement Learning

Authors: Le-Hung Hoang, Quang-Trung Luu, Dinh Thai Hoang, Diep N. Nguyen, Van-Dinh Nguyen

Abstract: Stacked intelligent metasurfaces (SIMs) have recently emerged as a powerful wave-domain technology that enables multi-stage manipulation of electromagnetic signals through multilayer programmable architectures. While SIMs offer unprecedented degrees of freedom for enhancing physical-layer security, their extremely large number of meta-atoms leads to a high-dimensional and strongly coupled optimiza… ▽ More Stacked intelligent metasurfaces (SIMs) have recently emerged as a powerful wave-domain technology that enables multi-stage manipulation of electromagnetic signals through multilayer programmable architectures. While SIMs offer unprecedented degrees of freedom for enhancing physical-layer security, their extremely large number of meta-atoms leads to a high-dimensional and strongly coupled optimization space, making conventional design approaches inefficient and difficult to scale. Moreover, existing deep reinforcement learning (DRL) techniques suffer from slow convergence and performance degradation in dynamic wireless environments with imperfect knowledge of passive eavesdroppers. To address these challenges, we propose a hybrid quantum proximal policy optimization (QPPO) framework for SIM-assisted secure communications that jointly optimizes transmit power allocation and SIM phase shifts to maximize the average secrecy rate under power and quality-of-service constraints. Specifically, a parameterized quantum circuit is embedded into the actor network, forming a hybrid classical-quantum policy architecture that enhances policy representation capability and exploration efficiency in high-dimensional continuous action spaces. Extensive simulations demonstrate that the proposed Q-PPO scheme consistently outperforms DRL baselines, achieving approximately 15% higher secrecy rates and 30% faster convergence under imperfect eavesdropper channel state information. These results establish Q-PPO as a powerful optimization paradigm for SIM-enabled secure wireless networks. △ Less

Submitted 28 May, 2026; v1 submitted 29 January, 2026; originally announced February 2026.

Comments: Submitted to IEEE TCOM: 13 pages

ViSpeechFormer: A Phonemic Approach for Vietnamese Automatic Speech Recognition

Authors: Khoa Anh Nguyen, Long Minh Hoang, Nghia Hieu Nguyen, Luan Thanh Nguyen, Ngan Luu-Thuy Nguyen

Abstract: Vietnamese has a phonetic orthography, where each grapheme corresponds to at most one phoneme and vice versa. Exploiting this high grapheme-phoneme transparency, we propose ViSpeechFormer (\textbf{Vi}etnamese \textbf{Speech} Trans\textbf{Former}), a phoneme-based approach for Vietnamese Automatic Speech Recognition (ASR). To the best of our knowledge, this is the first Vietnamese ASR framework tha… ▽ More Vietnamese has a phonetic orthography, where each grapheme corresponds to at most one phoneme and vice versa. Exploiting this high grapheme-phoneme transparency, we propose ViSpeechFormer (\textbf{Vi}etnamese \textbf{Speech} Trans\textbf{Former}), a phoneme-based approach for Vietnamese Automatic Speech Recognition (ASR). To the best of our knowledge, this is the first Vietnamese ASR framework that explicitly models phonemic representations. Experiments on two publicly available Vietnamese ASR datasets show that ViSpeechFormer achieves strong performance, generalizes better to out-of-vocabulary words, and is less affected by training bias. This phoneme-based paradigm is also promising for other languages with phonetic orthographies. The code will be released upon acceptance of this paper. △ Less

Submitted 10 February, 2026; originally announced February 2026.

The Benefits of Diversity: Combining Comparisons and Ratings for Efficient Scoring

Authors: Julien Fageot, Matthias Grossglauser, Lê-Nguyên Hoang, Matteo Tacchi-Bénard, Oscar Villemaud

Abstract: Should humans be asked to evaluate entities individually or comparatively? This question has been the subject of long debates. In this work, we show that, interestingly, combining both forms of preference elicitation can outperform the focus on a single kind. More specifically, we introduce SCoRa (Scoring from Comparisons and Ratings), a unified probabilistic model that allows to learn from both s… ▽ More Should humans be asked to evaluate entities individually or comparatively? This question has been the subject of long debates. In this work, we show that, interestingly, combining both forms of preference elicitation can outperform the focus on a single kind. More specifically, we introduce SCoRa (Scoring from Comparisons and Ratings), a unified probabilistic model that allows to learn from both signals. We prove that the MAP estimator of SCoRa is well-behaved. It verifies monotonicity and robustness guarantees. We then empirically show that SCoRa recovers accurate scores, even under model mismatch. Most interestingly, we identify a realistic setting where combining comparisons and ratings outperforms using either one alone, and when the accurate ordering of top entities is critical. Given the de facto availability of signals of multiple forms, SCoRa additionally offers a versatile foundation for preference learning. △ Less

Submitted 8 February, 2026; originally announced February 2026.

Comments: 1 table, 5 figures, 8 pages

Fine-Grained Frame Modeling in Multi-head Self-Attention for Speech Deepfake Detection

Authors: Tuan Dat Phuong, Duc-Tuan Truong, Long-Vu Hoang, Trang Nguyen Thi Thu

Abstract: Transformer-based models have shown strong performance in speech deepfake detection, largely due to the effectiveness of the multi-head self-attention (MHSA) mechanism. MHSA provides frame-level attention scores, which are particularly valuable because deepfake artifacts often occur in small, localized regions along the temporal dimension of speech. This makes fine-grained frame modeling essential… ▽ More Transformer-based models have shown strong performance in speech deepfake detection, largely due to the effectiveness of the multi-head self-attention (MHSA) mechanism. MHSA provides frame-level attention scores, which are particularly valuable because deepfake artifacts often occur in small, localized regions along the temporal dimension of speech. This makes fine-grained frame modeling essential for accurately detecting subtle spoofing cues. In this work, we propose fine-grained frame modeling (FGFM) for MHSA-based speech deepfake detection, where the most informative frames are first selected through a multi-head voting (MHV) module. These selected frames are then refined via a cross-layer refinement (CLR) module to enhance the model's ability to learn subtle spoofing cues. Experimental results demonstrate that our method outperforms the baseline model and achieves Equal Error Rate (EER) of 0.90%, 1.88%, and 6.64% on the LA21, DF21, and ITW datasets, respectively. These consistent improvements across multiple benchmarks highlight the effectiveness of our fine-grained modeling for robust speech deepfake detection. △ Less

Submitted 4 February, 2026; originally announced February 2026.

Comments: Accepted by ICASSP 2026

Photonic Reservoir Engineering via 2D $Λ$-Type Atomic Arrays in Waveguide QED

Authors: Thi Phuong Anh Nguyen, Le Phuong Hoang, Xuan Binh Cao

Abstract: Electromagnetically induced transparency (EIT) in $Λ$-type atomic systems underpins quantum technologies such as high-fidelity memory and nonlinear optics, but conventional setups face intrinsic limitations. Standard geometries of one-dimensional atomic chains coupled to waveguides allow only a single bright superradiant channel, while subradiant modes remain weakly accessible, limiting control ov… ▽ More Electromagnetically induced transparency (EIT) in $Λ$-type atomic systems underpins quantum technologies such as high-fidelity memory and nonlinear optics, but conventional setups face intrinsic limitations. Standard geometries of one-dimensional atomic chains coupled to waveguides allow only a single bright superradiant channel, while subradiant modes remain weakly accessible, limiting control over collective radiative behavior and dark-state pathways. This leads to unwanted inelastic processes, degrading memory fidelity and reducing nonlinear photon generation efficiency. Here, we propose two two-dimensional (2D) atomic lattice geometries coupled to a photonic crystal waveguide, namely Zigzag and Orthogonal structures. In the Zigzag model, engineered collective super- and subradiant modes produce a flattened EIT window, broadening the transmission bandwidth and suppressing unwanted scattering to enhance quantum memory fidelity. In the Orthogonal model, four-wave mixing (FWM) intensity is amplified by up to six orders of magnitude relative to a conventional one-dimensional $Λ$-type EIT chain with identical $Γ_{1D}$, $Ω_c$, and probe intensity, with localized idler photons forming well-defined spectral modes. These results demonstrate a versatile route to engineer structured photonic reservoirs for on-demand photon generation, high-fidelity quantum storage, and enhanced nonlinear optical processes. △ Less

Submitted 2 January, 2026; originally announced January 2026.

BLISS: Bandit Layer Importance Sampling Strategy for Efficient Training of Graph Neural Networks

Authors: Omar Alsaqa, Linh Thi Hoang, Muhammed Fatih Balin

Abstract: Graph Neural Networks (GNNs) are powerful tools for learning from graph-structured data, but their application to large graphs is hindered by computational costs. The need to process every neighbor for each node creates memory and computational bottlenecks. To address this, we introduce BLISS, a Bandit Layer Importance Sampling Strategy. It uses multi-armed bandits to dynamically select the most i… ▽ More Graph Neural Networks (GNNs) are powerful tools for learning from graph-structured data, but their application to large graphs is hindered by computational costs. The need to process every neighbor for each node creates memory and computational bottlenecks. To address this, we introduce BLISS, a Bandit Layer Importance Sampling Strategy. It uses multi-armed bandits to dynamically select the most informative nodes at each layer, balancing exploration and exploitation to ensure comprehensive graph coverage. Unlike existing static sampling methods, BLISS adapts to evolving node importance, leading to more informed node selection and improved performance. It demonstrates versatility by integrating with both Graph Convolutional Networks (GCNs) and Graph Attention Networks (GATs), adapting its selection policy to their specific aggregation mechanisms. Experiments show that BLISS maintains or exceeds the accuracy of full-batch training. △ Less

Submitted 26 December, 2025; originally announced December 2025.

Comments: Accepted for 5th Muslims in ML Workshop co-located with NeurIPS 2025. OpenReview: https://openreview.net/forum?id=VaHubA7Pwv Code: https://github.com/linhthi/BLISS-GNN

MSC Class: 68T05; 05C85; 62L05; 68T07 ACM Class: I.2.6; G.2.2; F.2.2

On compressible fluid flows of Forchheimer-type in rotating heterogeneous porous media

Authors: Emine Celik, Luan Hoang, Thinh Kieu

Abstract: We study the dynamics of compressible fluids in rotating heterogeneous porous media. The fluid flow is of {F}orchheimer-type and is subject to a mixed mass and volumetric flux boundary condition. The governing equations are reduced to a nonlinear partial differential equation for the pseudo-pressure. This parabolic-typed equation can be degenerate and/or singular in the spatial variables, the unkn… ▽ More We study the dynamics of compressible fluids in rotating heterogeneous porous media. The fluid flow is of {F}orchheimer-type and is subject to a mixed mass and volumetric flux boundary condition. The governing equations are reduced to a nonlinear partial differential equation for the pseudo-pressure. This parabolic-typed equation can be degenerate and/or singular in the spatial variables, the unknown and its gradient. We establish the $L^α$-estimate for the solutions, for any positive number $α$, in terms of the initial and boundary data and the angular speed of rotation. It requires new elliptic and parabolic Sobolev inequalities and trace theorem with multiple weights that are suitable to the nonlinear structure of the equation. The $L^\infty$-estimate is then obtained without imposing any conditions on the $L^\infty$-norms of the weights and the initial and boundary data. △ Less

Submitted 25 May, 2026; v1 submitted 15 December, 2025; originally announced December 2025.

Comments: 47 pages, improved presentation. to appear in Mathematical Methods in the Applied Sciences

MSC Class: 35Q35; 76S05; 35B45; 35K20; 35K65; 35K67

Collective Magnetic Excitations in a Photo-excited Electron-doped Cuprate Superconductor

Authors: Daniel Jost, Jiarui Li, Jordyn Hales, Jonathan Sobota, Giacomo Merzoni, Leonardo Martinelli, Shuhan Ding, Kejun Xu, Justine Schlappa, Andreas Scherz, Robert Carley, Benjamin E. Van Kuiken, Teguh C. Asmara, Le Phuong Hoang, Laurent Mercadier, Sergii Parchenko, Martin Teichmann, Patrick S. Kirchmann, Giacomo Ghiringhelli, Brian Moritz, Zhi-Xun Shen, Thomas P. Devereaux, Yao Wang, Wei-Sheng Lee

Abstract: Elucidating the microscopic behavior of cuprates under ultrafast photoexcitation offers critical insights into their highly correlated out-of-equilibrium states. Although quasiparticle dynamics have been investigated extensively, the behavior of collective magnetic excitations remains comparatively unexplored. Here, we use time-resolved resonant inelastic X-ray scattering (trRIXS) at the Cu $L_3$-… ▽ More Elucidating the microscopic behavior of cuprates under ultrafast photoexcitation offers critical insights into their highly correlated out-of-equilibrium states. Although quasiparticle dynamics have been investigated extensively, the behavior of collective magnetic excitations remains comparatively unexplored. Here, we use time-resolved resonant inelastic X-ray scattering (trRIXS) at the Cu $L_3$-edge to track the collective magnetic excitations (paramagnons) in an optimally electron-doped cuprate driven out-of-equilibrium by a femtosecond pump laser pulse. Upon pumping, we observed an anti-Stokes signal associated with paramagnon generation, which modifies the paramagnon dispersion near the zone center, although the bandwidth remained unchanged. Moreover, the spectral weight exhibits a momentum-dependent variation across the Brillouin zone. The light-driven boost of the paramagnon population and the resulting spectral-weight transfer could provide new leverage to manipulate the properties of cuprates. △ Less

Submitted 14 July, 2026; v1 submitted 26 November, 2025; originally announced November 2025.

Comments: Accepted in Physical Review Letters

High frame rate RIXS spectroscopy using a JUNGFRAU detector with an iLGAD sensor

Authors: Nuno Duarte, Loïc Le Guyader, Viktoria Hinger, Marco Ramilli, Justine Schlappa, Robert Carley, Maria Carulla, Yi-Ping Chang, Devesh Chopra, Natalia Gerasimova, Michael Grimes, Aldo Mozzanica, Sergii Parchenko, Le Phuong Hoang, Urs Staub, Jiaguo Zhang, Anna Bergamaschi, Andreas Scherz, Bernd Schmitt, Monica Turcato

Abstract: Resonant inelastic X-ray scattering (RIXS) is a powerful photon-in, photon-out spectroscopy technique for probing electronic, magnetic, and lattice excitations in matter. Time-resolved RIXS extends this capability through a stroboscopic optical pump-probe scheme to characterize the time evolution of the photoexcitation and subsequent relaxation dynamics of a sample. This technique is, however, ext… ▽ More Resonant inelastic X-ray scattering (RIXS) is a powerful photon-in, photon-out spectroscopy technique for probing electronic, magnetic, and lattice excitations in matter. Time-resolved RIXS extends this capability through a stroboscopic optical pump-probe scheme to characterize the time evolution of the photoexcitation and subsequent relaxation dynamics of a sample. This technique is, however, extremely photon-hungry, requiring high-repetition-rate and intense X-ray facilities. The Heisenberg RIXS (hRIXS) spectrometer at the Spectroscopy and Coherent Scattering (SCS) instrument of the European X-ray Free-Electron Laser (EuXFEL) is designed to exploit high-repetition-rates, while maintaining optimal time and energy resolution. In this work, we demonstrate the successful deployment of a JUNGFRAU detector equipped with an inverse Low Gain Avalanche Diode (iLGAD) sensor for time-resolved RIXS studies in the soft X-ray range, using the hRIXS spectrometer. A spatial resolution of $19.71 \pm 0.7~μ\mathrm{m}$ and a resolving power exceeding 10,000 were achieved at an unprecedented frame rate of 47 kHz. Intra-train resolved data measured with a high FEL peak fluence of $1.8~\mathrm{mJ/cm^{2}}$ for a 928.5 eV ph photon energy and 1.1 MHz repetition rate from cupric oxide (CuO) revealed a decrease in the emitted signal by ~10% over a time interval of $340~μ\mathrm{s}$, indicating FEL-induced effects that require monitoring when conducting high-repetition-rate experiments. These results establish the JUNGFRAU-iLGAD as a promising detector to harvest the full potential of the hRIXS spectrometer, and validate its suitability for soft X-ray applications. △ Less

Submitted 15 November, 2025; originally announced November 2025.

Comments: 30 pages, 13 figures

Linear non-divergence elliptic equations in a bounded, infinitely winding planar domain

Authors: Luan Hoang, Akif Ibragimov

Abstract: We study the second order elliptic equations of non-divergence form in a planar domain with complicated geometry. In this case the domain winds around a fixed circle infinitely many times and converges to it when the rotating angle goes to infinity. For the homogeneous equation and the homogeneous Dirichlet boundary condition, in the case of bounded drifts, we prove that the maximum of the solutio… ▽ More We study the second order elliptic equations of non-divergence form in a planar domain with complicated geometry. In this case the domain winds around a fixed circle infinitely many times and converges to it when the rotating angle goes to infinity. For the homogeneous equation and the homogeneous Dirichlet boundary condition, in the case of bounded drifts, we prove that the maximum of the solution on the cross-section corresponding to a given rotating angle either grows or decays exponentially as the angle goes to infinity. Results for the oscillation and its asymptotic estimates are also obtained for inhomogeneous Dirichlet data. If the drift is unbounded but does not grow to infinity too fast, then the above maximum also goes to either zero or infinity. For the inhomogeneous equation, we obtain the estimates in the case of bounded forcing functions. Moreover, we establish the uniqueness of the solution and its continuous dependence on the boundary data and the forcing function. △ Less

Submitted 17 February, 2026; v1 submitted 17 September, 2025; originally announced September 2025.

Comments: minor fixes. to appear in Complex Variables and Elliptic Equations, 31 pages

MSC Class: 35A09; 35B30; 35B40; 35B50

Scaling Nanoribbon Transistors with Monolayer Transition Metal Dichalcogenides

Authors: Tara Peña, Anton E. O. Persson, Andrey Krayev, Áshildur Friðriksdóttir, Haotian Su, Yuan-Mau Lee, Young Suh Song, Kathryn Neilson, Zhepeng Zhang, Anh Tuan Hoang, Jerry A. Yang, Lauren Hoang, Shan X. Wang, Andrew J. Mannix, Paul C. McIntyre, Eric Pop

Abstract: Nanoscale transistors require aggressive reduction of all channel dimensions: length, width, and thickness. While monolayer two-dimensional semiconductors (2DS) offer ultimate thickness scaling, good performance has largely been achieved only in micrometer-wide channels. Here, we demonstrate both $\it{n}$- and $\it{p}$-type nanoribbon transistors based on monolayer 2DS, fabricated using a multi-pa… ▽ More Nanoscale transistors require aggressive reduction of all channel dimensions: length, width, and thickness. While monolayer two-dimensional semiconductors (2DS) offer ultimate thickness scaling, good performance has largely been achieved only in micrometer-wide channels. Here, we demonstrate both $\it{n}$- and $\it{p}$-type nanoribbon transistors based on monolayer 2DS, fabricated using a multi-patterning process, reaching channel widths and lengths down to 25-30 nm. 'Anchored' contacts improve device yield, while nanoscale imaging, including tip-enhanced photoluminescence, reveals minimal edge degradation. The devices reach on-state currents up to 560, 420, and 130 $μ$A $μ$m$^{-1}$ at 1 V drain-to-source voltage for $\it{n}$-type MoS$_{2}$, WS$_{2}$, and $\it{p}$-type WSe$_{2}$, respectively, integrated with thin high-$κ$ dielectrics. These results surpass prior reports for single-gated nanoribbons, the WS$_{2}$ by over 100 times, even in normally off (enhancement-mode) transistors. Taken together, these findings suggest that top down patterned 2DS nanoribbons are promising building blocks for future nanosheet transistors. △ Less

Submitted 4 June, 2026; v1 submitted 12 September, 2025; originally announced September 2025.

Comments: 22 pages, 9 figures

Journal ref: Nature Nanotechnology 21 (2026) 803-809

Investigating Student Interaction Patterns with Large Language Model-Powered Course Assistants in Computer Science Courses

Authors: Chang Liu, Loc Hoang, Andrew Stolman, Rene F. Kizilcec, Bo Wu

Abstract: Providing students with flexible and timely academic support is a challenge at most colleges and universities, leaving many students without help outside scheduled hours. Large language models (LLMs) are promising for bridging this gap, but interactions between students and LLMs are rarely overseen by educators. We developed and studied an LLM-powered course assistant deployed across multiple comp… ▽ More Providing students with flexible and timely academic support is a challenge at most colleges and universities, leaving many students without help outside scheduled hours. Large language models (LLMs) are promising for bridging this gap, but interactions between students and LLMs are rarely overseen by educators. We developed and studied an LLM-powered course assistant deployed across multiple computer science courses to characterize real-world use and understand pedagogical implications. By Spring 2024, our system had been deployed to approximately 2,000 students across six courses at three institutions. Analysis of the interaction data shows that usage remains strong in the evenings and nights and is higher in introductory courses, indicating that our system helps address temporal support gaps and novice learner needs. We sampled 200 conversations per course for manual annotation: most sampled responses were judged correct and helpful, with a small share unhelpful or erroneous; few responses included dedicated examples. We also examined an inquiry-based learning strategy: only around 11% of sampled conversations contained LLM-generated follow-up questions, which were often ignored by students in advanced courses. A Bloom's taxonomy analysis reveals that current LLM capabilities are limited in generating higher-order cognitive questions. These patterns suggest opportunities for pedagogically oriented LLM-based educational systems and greater educator involvement in configuring prompts, content, and policies. △ Less

Submitted 9 September, 2025; originally announced September 2025.

Deep Learning to Automate Parameter Extraction and Model Fitting of Two-Dimensional Transistors

Authors: Robert K. A. Bennett, Jan-Lucas Uslu, Harmon F. Gault, Asir Intisar Khan, Lauren Hoang, Tara Peña, Kathryn Neilson, Young Suh Song, Zhepeng Zhang, Andrew J. Mannix, Eric Pop

Abstract: We present a deep learning approach to extract physical parameters (e.g., mobility, Schottky contact barrier height, defect profiles) of two-dimensional (2D) transistors from electrical measurements, enabling automated parameter extraction and technology computer-aided design (TCAD) fitting. To facilitate this task, we implement a simple data augmentation and pre-training approach by training a se… ▽ More We present a deep learning approach to extract physical parameters (e.g., mobility, Schottky contact barrier height, defect profiles) of two-dimensional (2D) transistors from electrical measurements, enabling automated parameter extraction and technology computer-aided design (TCAD) fitting. To facilitate this task, we implement a simple data augmentation and pre-training approach by training a secondary neural network to approximate a physics-based device simulator. This method enables high-quality fits after training the neural network on electrical data generated from physics-based simulations of ~500 devices, a factor >40$\times$ fewer than other recent efforts. Consequently, fitting can be achieved by training on physically rigorous TCAD models, including complex geometry, self-consistent transport, and electrostatic effects, and is not limited to computationally inexpensive compact models. We apply our approach to reverse-engineer key parameters from experimental monolayer WS$_2$ transistors, achieving a median coefficient of determination ($R^2$) = 0.99 when fitting measured electrical data. We also demonstrate that this approach generalizes and scales well by reverse-engineering electrical data on high-electron-mobility transistors while fitting 35 parameters simultaneously. To facilitate future research on deep learning approaches for inverse transistor design, we have published our code and sample data sets online. △ Less

Submitted 7 July, 2025; originally announced July 2025.

Comments: Main text + supplementary information

Asymptotic estimates for solutions of inhomogeneous non-divergence diffusion equations with drifts

Authors: Luan Hoang, Akif Ibragimov

Abstract: We study the long-time dynamics of the nonlinear processes modeled by diffusion-transport partial differential equations in non-divergence form with drifts. The solutions are subject to some inhomogeneous Dirichlet boundary condition. Starting with the reduced linear problem, we obtain the asymptotic estimates for the solutions, as time $t\to\infty$, depending on the asymptotic behavior of the for… ▽ More We study the long-time dynamics of the nonlinear processes modeled by diffusion-transport partial differential equations in non-divergence form with drifts. The solutions are subject to some inhomogeneous Dirichlet boundary condition. Starting with the reduced linear problem, we obtain the asymptotic estimates for the solutions, as time $t\to\infty$, depending on the asymptotic behavior of the forcing term and boundary data. These are established in both cases when the drifts are uniformly bounded, and unbounded as $t\to\infty$. For the nonlinear problem, we prove the convergence of the solutions under suitable conditions that balance the growth of the nonlinear term with the decay of the data. To take advantage of the diffusion in the non-divergence form, we prove an inhomogeneous version of the Landis-typed Growth Lemma and apply it to successive time-intervals. At each time step, the center for the barrier function is selected carefully to optimize the contracting factor. Our rigorous results show the robustness of the model. △ Less

Submitted 10 February, 2026; v1 submitted 23 June, 2025; originally announced June 2025.

Comments: minor fixes. to appear in Journal of Mathematical Analysis and Applications (34 pages)

MSC Class: 35Q35; 35A09; 35B50; 35B40

Pushing the Performance of Synthetic Speech Detection with Kolmogorov-Arnold Networks and Self-Supervised Learning Models

Authors: Tuan Dat Phuong, Long-Vu Hoang, Huy Dat Tran

Abstract: Recent advancements in speech synthesis technologies have led to increasingly advanced spoofing attacks, posing significant challenges for automatic speaker verification systems. While systems based on self-supervised learning (SSL) models, particularly the XLSR-Conformer model, have demonstrated remarkable performance in synthetic speech detection, there remains room for architectural improvement… ▽ More Recent advancements in speech synthesis technologies have led to increasingly advanced spoofing attacks, posing significant challenges for automatic speaker verification systems. While systems based on self-supervised learning (SSL) models, particularly the XLSR-Conformer model, have demonstrated remarkable performance in synthetic speech detection, there remains room for architectural improvements. In this paper, we propose a novel approach that replaces the traditional Multi-Layer Perceptron in the XLSR-Conformer model with a Kolmogorov-Arnold Network (KAN), a novel architecture based on the Kolmogorov-Arnold representation theorem. Our results on ASVspoof2021 demonstrate that integrating KAN into the SSL-based models can improve the performance by 60.55% relatively on LA and DF sets, further achieving 0.70% EER on the 21LA set. These findings suggest that incorporating KAN into SSL-based models is a promising direction for advances in synthetic speech detection. △ Less

Submitted 16 June, 2025; originally announced June 2025.

Comments: Accepted to Interspeech 2025

Acoustic scattering AI for non-invasive object classifications: A case study on hair assessment

Authors: Long-Vu Hoang, Tuan Nguyen, Tran Huy Dat

Abstract: This paper presents a novel non-invasive object classification approach using acoustic scattering, demonstrated through a case study on hair assessment. When an incident wave interacts with an object, it generates a scattered acoustic field encoding structural and material properties. By emitting acoustic stimuli and capturing the scattered signals from head-with-hair-sample objects, we classify h… ▽ More This paper presents a novel non-invasive object classification approach using acoustic scattering, demonstrated through a case study on hair assessment. When an incident wave interacts with an object, it generates a scattered acoustic field encoding structural and material properties. By emitting acoustic stimuli and capturing the scattered signals from head-with-hair-sample objects, we classify hair type and moisture using AI-driven, deep-learning-based sound classification. We benchmark comprehensive methods, including (i) fully supervised deep learning, (ii) embedding-based classification, (iii) supervised foundation model fine-tuning, and (iv) self-supervised model fine-tuning. Our best strategy achieves nearly 90% classification accuracy by fine-tuning all parameters of a self-supervised model. These results highlight acoustic scattering as a privacy-preserving, non-contact alternative to visual classification, opening huge potential for applications in various industries. △ Less

Submitted 19 May, 2026; v1 submitted 16 June, 2025; originally announced June 2025.

Comments: This paper has been retracted by the authors. Due to miscommunication, the authorship is incomplete and missing early contributions

Qwen vs. Gemma Integration with Whisper: A Comparative Study in Multilingual SpeechLLM Systems

Authors: Tuan Nguyen, Long-Vu Hoang, Huy-Dat Tran

Abstract: This paper presents our system for the MLC-SLM Challenge 2025, focusing on multilingual speech recognition and language modeling with large language models (LLMs). Our approach combines a fine-tuned Whisper-large-v3 encoder with efficient projector architectures and various decoder configurations. We employ a three-stage training methodology that progressively optimizes the encoder, projector, and… ▽ More This paper presents our system for the MLC-SLM Challenge 2025, focusing on multilingual speech recognition and language modeling with large language models (LLMs). Our approach combines a fine-tuned Whisper-large-v3 encoder with efficient projector architectures and various decoder configurations. We employ a three-stage training methodology that progressively optimizes the encoder, projector, and LLM components. Our system achieves competitive performance with a private test average WER/CER result of 16.63% using the Gemma3-12B and 18.6% using the Qwen2.5-7B as decoder-only language model. △ Less

Submitted 7 July, 2025; v1 submitted 16 June, 2025; originally announced June 2025.

Comments: Accepted to Interspeech MLCSLM-2025 Workshop

On Monotonicity in AI Alignment

Authors: Gilles Bareilles, Julien Fageot, Lê-Nguyên Hoang, Peva Blanchard, Wassim Bouaziz, Sébastien Rouault, El-Mahdi El-Mhamdi

Abstract: Comparison-based preference learning has become central to the alignment of AI models with human preferences. However, these methods may behave counterintuitively. After empirically observing that, when accounting for a preference for response $y$ over $z$, the model may actually decrease the probability (and reward) of generating $y$ (an observation also made by others), this paper investigates t… ▽ More Comparison-based preference learning has become central to the alignment of AI models with human preferences. However, these methods may behave counterintuitively. After empirically observing that, when accounting for a preference for response $y$ over $z$, the model may actually decrease the probability (and reward) of generating $y$ (an observation also made by others), this paper investigates the root causes of (non) monotonicity, for a general comparison-based preference learning framework that subsumes Direct Preference Optimization (DPO), Generalized Preference Optimization (GPO) and Generalized Bradley-Terry (GBT). Under mild assumptions, we prove that such methods still satisfy what we call local pairwise monotonicity. We also provide a bouquet of formalizations of monotonicity, and identify sufficient conditions for their guarantee, thereby providing a toolbox to evaluate how prone learning models are to monotonicity violations. These results clarify the limitations of current methods and provide guidance for developing more trustworthy preference learning algorithms. △ Less

Submitted 10 June, 2025; originally announced June 2025.

Generalizing while preserving monotonicity in comparison-based preference learning models

Authors: Julien Fageot, Peva Blanchard, Gilles Bareilles, Lê-Nguyên Hoang

Abstract: If you tell a learning model that you prefer an alternative $a$ over another alternative $b$, then you probably expect the model to be monotone, that is, the valuation of $a$ increases, and that of $b$ decreases. Yet, perhaps surprisingly, many widely deployed comparison-based preference learning models, including large language models, fail to have this guarantee. Until now, the only comparison-b… ▽ More If you tell a learning model that you prefer an alternative $a$ over another alternative $b$, then you probably expect the model to be monotone, that is, the valuation of $a$ increases, and that of $b$ decreases. Yet, perhaps surprisingly, many widely deployed comparison-based preference learning models, including large language models, fail to have this guarantee. Until now, the only comparison-based preference learning algorithms that were proved to be monotone are the Generalized Bradley-Terry models. Yet, these models are unable to generalize to uncompared data. In this paper, we advance the understanding of the set of models with generalization ability that are monotone. Namely, we propose a new class of Linear Generalized Bradley-Terry models with Diffusion Priors, and identify sufficient conditions on alternatives' embeddings that guarantee monotonicity. Our experiments show that this monotonicity is far from being a general guarantee, and that our new class of generalizing models improves accuracy, especially when the dataset is limited. △ Less

Submitted 22 October, 2025; v1 submitted 10 June, 2025; originally announced June 2025.

Comments: Accepted at Neurips 2025

A priori estimates for gaseous flows of Forchheimer-type in heterogeneous porous media

Authors: Emine Celik, Luan Hoang, Thinh Kieu

Abstract: We study isentropic fluid flows of gases of the Forchheimer-type in heterogeneous porous media. The governing equation is a doubly nonlinear parabolic equation with coefficients depending on the spatial variables. Its solutions are subject to a nonlinear Robin boundary condition. We establish the estimates of the solutions for short time in terms of the initial and boundary data. For the proof, th… ▽ More We study isentropic fluid flows of gases of the Forchheimer-type in heterogeneous porous media. The governing equation is a doubly nonlinear parabolic equation with coefficients depending on the spatial variables. Its solutions are subject to a nonlinear Robin boundary condition. We establish the estimates of the solutions for short time in terms of the initial and boundary data. For the proof, the multi-weight versions of the Sobolev inequality, parabolic Sobolev inequality and trace theorem are derived. They are then used to implement the Moser iteration for suitable weighted norms. △ Less

Submitted 17 May, 2025; originally announced May 2025.

Comments: submitted for publication. 43 pages

MSC Class: 35Q35; 76S05; 35B45; 35G31; 35M13

Low Resistance P-type Contacts to Monolayer WSe$_2$ through Chlorinated Solvent Doping

Authors: Lauren Hoang, Robert K. A. Bennett, Anh Tuan Hoang, Tara Pena, Zhepeng Zhang, Marisa Hocking, Ashley P. Saunders, Fang Liu, Eric Pop, Andrew J. Mannix

Abstract: Tungsten diselenide (WSe$_2$) is a promising p-type semiconductor limited by high contact resistance ($R_\textrm{C}$) and the lack of a reliable doping strategy. Here, we demonstrate that exposing WSe$_2$ to chloroform provides simple and stable p-type doping. In monolayer WSe$_2$ transistors with Pd contacts, chloroform increases the maximum hole current by over 100$\times$ (>200 $μ$A/$μ$m), redu… ▽ More Tungsten diselenide (WSe$_2$) is a promising p-type semiconductor limited by high contact resistance ($R_\textrm{C}$) and the lack of a reliable doping strategy. Here, we demonstrate that exposing WSe$_2$ to chloroform provides simple and stable p-type doping. In monolayer WSe$_2$ transistors with Pd contacts, chloroform increases the maximum hole current by over 100$\times$ (>200 $μ$A/$μ$m), reduces $R_\textrm{C}$ to ~2.5 k$Ω\cdotμ$m, and retains an on/off ratio of $10^{10}$ at room temperature. These improvements persist for over 8 months, survive annealing above 150 °C, and remain effective down to 10 K, enabling a cryogenic $R_\textrm{C}$ of ~1 k$Ω\cdotμ$m. Density functional theory indicates that chloroform strongly physisorbs to WSe$_2$, inducing hole doping with minimal impact on the electronic states between the valence band and conduction band edges. Auger electron spectroscopy and atomic force microscopy reveal that chloroform intercalates at the WSe$_2$ interface with the gate oxide, contributing to doping stability and mitigating interfacial dielectric disorder. This robust, scalable approach enables high-yield WSe$_2$ transistors with good p-type performance. △ Less

Submitted 29 April, 2025; originally announced April 2025.

Synthesis-related nanoscale defects in Mo-based Janus monolayers revealed by cross-correlated AFM and TERS imaging

Authors: Tianyi Zhang, Andrey Krayev, Tilo H. Yang, Nannan Mao, Lauren Hoang, Zhien Wang, Hongwei Liu, Yu-Ren Peng, Yunyue Zhu, Eleonora Isotta, Maria E. Kira, Ariete Righi, Marcos A. Pimenta, Yu-Lun Chueh, Eric Pop, Andrew J. Mannix, Jing Kong

Abstract: Two-dimensional (2D) Janus transition metal dichalcogenides (TMDs) are promising candidates for various applications in non-linear optics, energy harvesting, and catalysis. These materials are usually synthesized via chemical conversion of pristine TMDs. Nanometer-scale characterization of the obtained Janus materials' morphology and local composition is crucial for both the synthesis optimization… ▽ More Two-dimensional (2D) Janus transition metal dichalcogenides (TMDs) are promising candidates for various applications in non-linear optics, energy harvesting, and catalysis. These materials are usually synthesized via chemical conversion of pristine TMDs. Nanometer-scale characterization of the obtained Janus materials' morphology and local composition is crucial for both the synthesis optimization and the future device applications. In this work, we present a cross-correlated atomic force microscopy (AFM) and tip-enhanced Raman spectroscopy (TERS) study of Janus $\mathrm{Mo}_{\mathrm{Se}}^{\mathrm{S}}$ and Janus $\mathrm{Mo}_{\mathrm{S}}^{\mathrm{Se}}$ monolayers synthesized by the hydrogen plasma-assisted chemical conversion of $\mathrm{MoSe}_2$ and $\mathrm{MoS}_2$, respectively. We demonstrate how the choice of the growth substrate and the starting TMD affects the morphology of the resulting Janus material. Furthermore, by employing TERS imaging, we demonstrate the presence of nanoscale islands (~20 nm across) of $\mathrm{MoSe}_2$-$\mathrm{Mo}_{\mathrm{Se}}^{\mathrm{S}}$ ($\mathrm{MoS}_2$-$\mathrm{Mo}_{\mathrm{S}}^{\mathrm{Se}}$) vertical heterostructures originating from the bilayer nanoislands in the precursor monolayer crystals. The understanding of the origins of nanoscale defects in Janus TMDs revealed in our study can help with further optimization of the Janus conversion process towards uniform and wrinkle-/crack-free Janus materials. Moreover, our work shows that cross-correlated AFM and TERS imaging is a powerful and accessible method for studying nanoscale composition and defects in Janus TMD monolayers. △ Less

Submitted 28 March, 2025; originally announced March 2025.

Ultrafast decoupling of polarization and strain in ferroelectric BaTiO$_3$

Authors: Le Phuong Hoang, David Pesquera, Gerard N. Hinsley, Robert Carley, Laurent Mercadier, Martin Teichmann, Saptam Ganguly, Teguh Citra Asmara, Giacomo Merzoni, Sergii Parchenko, Justine Schlappa, Zhong Yin, José Manuel Caicedo Roque, José Santiso, Irena Spasojevic, Cammille Carinan, Tien-Lin Lee, Kai Rossnage, Jörg Zegenhagen, Gustau Catalan, Ivan A. Vartanyants, Andreas Scherz, Giuseppe Mercurio

Abstract: A fundamental understanding of the interplay between lattice structure, polarization and electrons is pivotal to the optical control of ferroelectrics. The interaction between light and matter enables the remote and wireless control of the ferroelectric polarization on the picosecond timescale, while inducing strain, i.e., lattice deformation. At equilibrium, the ferroelectric polarization is prop… ▽ More A fundamental understanding of the interplay between lattice structure, polarization and electrons is pivotal to the optical control of ferroelectrics. The interaction between light and matter enables the remote and wireless control of the ferroelectric polarization on the picosecond timescale, while inducing strain, i.e., lattice deformation. At equilibrium, the ferroelectric polarization is proportional to the strain, and is typically assumed to be so also out of equilibrium. Decoupling the polarization from the strain would remove the constraint of sample design and provide an effective knob to manipulate the polarization by light. Here, upon an above-bandgap laser excitation of the prototypical ferroelectric BaTiO$_3$, we induce and measure an ultrafast decoupling between polarization and strain that begins within 350 fs, by softening Ti-O bonds via charge transfer, and lasts for several tens of picoseconds. We show that the ferroelectric polarization out of equilibrium is mainly determined by photoexcited electrons, instead of the strain. This excited state could serve as a starting point to achieve stable and reversible polarization switching via THz light. Our results demonstrate a light-induced transient and reversible control of the ferroelectric polarization and offer a pathway to control by light both electric and magnetic degrees of freedom in multiferroics. △ Less

Submitted 25 March, 2025; originally announced March 2025.

Journal ref: Nat. Commun. 16 (2025) 7966

The Science of the Einstein Telescope

Authors: Adrian Abac, Raul Abramo, Simone Albanesi, Angelica Albertini, Alessandro Agapito, Michalis Agathos, Conrado Albertus, Nils Andersson, Tomas Andrade, Igor Andreoni, Federico Angeloni, Marco Antonelli, John Antoniadis, Fabio Antonini, Manuel Arca Sedda, M. Celeste Artale, Stefano Ascenzi, Pierre Auclair, Matteo Bachetti, Charles Badger, Biswajit Banerjee, David Barba-Gonzalez, Daniel Barta, Nicola Bartolo, Andreas Bauswein , et al. (463 additional authors not shown)

Abstract: Einstein Telescope (ET) is the European project for a gravitational-wave (GW) observatory of third-generation. In this paper we present a comprehensive discussion of its science objectives, providing state-of-the-art predictions for the capabilities of ET in both geometries currently under consideration, a single-site triangular configuration or two L-shaped detectors. We discuss the impact that E… ▽ More Einstein Telescope (ET) is the European project for a gravitational-wave (GW) observatory of third-generation. In this paper we present a comprehensive discussion of its science objectives, providing state-of-the-art predictions for the capabilities of ET in both geometries currently under consideration, a single-site triangular configuration or two L-shaped detectors. We discuss the impact that ET will have on domains as broad and diverse as fundamental physics, cosmology, early Universe, astrophysics of compact objects, physics of matter in extreme conditions, and dynamics of stellar collapse. We discuss how the study of extreme astrophysical events will be enhanced by multi-messenger observations. We highlight the ET synergies with ground-based and space-borne GW observatories, including multi-band investigations of the same sources, improved parameter estimation, and complementary information on astrophysical or cosmological mechanisms obtained combining observations from different frequency bands. We present advancements in waveform modeling dedicated to third-generation observatories, along with open tools developed within the ET Collaboration for assessing the scientific potentials of different detector configurations. We finally discuss the data analysis challenges posed by third-generation observatories, which will enable access to large populations of sources and provide unprecedented precision. △ Less

Submitted 29 August, 2025; v1 submitted 15 March, 2025; originally announced March 2025.

Comments: 899 pages, 205 figures, v2: minor improvements, the version to appear in JCAP

Report number: ET-0036E-25

Journal ref: JCAP 03 (2026) 081

A Case for Specialisation in Non-Human Entities

Authors: El-Mahdi El-Mhamdi, Lê-Nguyên Hoang, Mariame Tighanimine

Abstract: With the rise of large multi-modal AI models, fuelled by recent interest in large language models (LLMs), the notion of artificial general intelligence (AGI) went from being restricted to a fringe community, to dominate mainstream large AI development programs. In contrast, in this paper, we make a case for specialisation, by reviewing the pitfalls of generality and stressing the industrial value… ▽ More With the rise of large multi-modal AI models, fuelled by recent interest in large language models (LLMs), the notion of artificial general intelligence (AGI) went from being restricted to a fringe community, to dominate mainstream large AI development programs. In contrast, in this paper, we make a case for specialisation, by reviewing the pitfalls of generality and stressing the industrial value of specialised systems. Our contribution is threefold. First, we review the most widely accepted arguments against specialisation, and discuss how their relevance in the context of human labour is actually an argument for specialisation in the case of non human agents, be they algorithms or human organisations. Second, we propose four arguments in favor of specialisation, ranging from machine learning robustness, to computer security, social sciences and cultural evolution. Third, we finally make a case for specification, discuss how the machine learning approach to AI has so far failed to catch up with good practices from safety-engineering and formal verification of software, and discuss how some emerging good practices in machine learning help reduce this gap. In particular, we justify the need for specified governance for hard-to-specify systems. △ Less

Submitted 21 August, 2025; v1 submitted 5 February, 2025; originally announced March 2025.

Comments: Accepted to AAAI/ACM AIES 2025

A class of anisotropic diffusion-transport equations in non-divergence form

Authors: Luan Hoang, Akif Ibragimov

Abstract: We generalize Einstein's probabilistic method for the Brownian motion to study compressible fluids in porous media. The multi-dimensional case is considered with general probability distribution functions. By relating the expected displacement per unit time with the velocity of the fluid, we derive an anisotropic diffusion equation in non-divergence form that contains a transport term. Under the D… ▽ More We generalize Einstein's probabilistic method for the Brownian motion to study compressible fluids in porous media. The multi-dimensional case is considered with general probability distribution functions. By relating the expected displacement per unit time with the velocity of the fluid, we derive an anisotropic diffusion equation in non-divergence form that contains a transport term. Under the Darcy law assumption, a corresponding nonlinear partial differential equations for the density function is obtained. The classical solutions of this equation are studied, and the maximum and strong maximum principles are established. We also obtain exponential decay estimates for the solutions for all time, and particularly, their exponential convergence as time tends to infinity. Our analysis uses some transformations of the Bernstein-Cole--Hopf type which are explicitly constructed even for very general equation of state. Moreover, the Lemma of Growth in time is proved and utilized in order to achieve the above decaying estimates. △ Less

Submitted 4 March, 2025; originally announced March 2025.

Comments: 24 pp, submitted for publication

MSC Class: 35Q35; 35Q86; 35A09; 35B50; 35B40

Resolving the Electron Plume within a Scanning Electron Microscope

Authors: Francis M. Alcorn, Christopher Perez, Eric J. Smoll, Lauren Hoang, Frederick Nitta, Andrew J. Mannix, A. Alec Talin, Craig Y. Nakakura, David W. Chandler, Suhas Kumar

Abstract: Scanning electron microscopy (SEM), a century-old technique, is today a ubiquitous method of imaging the surface of nanostructures. However, most SEM detectors simply count the number of secondary electrons from a material of interest, and thereby overlook the rich material information contained within them. Here, by simple modifications to a standard SEM tool, we resolve the momentum and energy i… ▽ More Scanning electron microscopy (SEM), a century-old technique, is today a ubiquitous method of imaging the surface of nanostructures. However, most SEM detectors simply count the number of secondary electrons from a material of interest, and thereby overlook the rich material information contained within them. Here, by simple modifications to a standard SEM tool, we resolve the momentum and energy information of secondary electrons by directly imaging the electron plume generated by the electron beam of the SEM. Leveraging these spectroscopic imaging capabilities, our technique is able to image lateral electric fields across a prototypical silicon p-n junctions and to distinguish differently doped regions, even when buried beyond depths typically accessible by SEM. Intriguingly, the sub-surface sensitivity of this technique reveals unexpectedly strong surface band bending within nominally passivated semiconductor structures, providing useful insights for complex layered component designs, in which interfacial dynamics dictate device operation. These capabilities for non-invasive, multi-modal probing of complicated electronic components are crucial in today's electronic manufacturing but is largely inaccessible even with sophisticated techniques. These results show that seemingly simple SEM can be extended to probe complex and useful material properties. △ Less

Submitted 10 January, 2025; originally announced January 2025.

Comments: Main text: 16 pages, 5 figure. Supporting information: 18 pages, 14 figures

Journal ref: ACS Nano 2024, 18, 49, 33479-33490

Optimizing Multi-Stage Language Models for Effective Text Retrieval

Authors: Quang Hoang Trung, Le Trung Hoang, Nguyen Van Hoang Phuc

Abstract: Efficient text retrieval is critical for applications such as legal document analysis, particularly in specialized contexts like Japanese legal systems. Existing retrieval methods often underperform in such domain-specific scenarios, necessitating tailored approaches. In this paper, we introduce a novel two-phase text retrieval pipeline optimized for Japanese legal datasets. Our method leverages a… ▽ More Efficient text retrieval is critical for applications such as legal document analysis, particularly in specialized contexts like Japanese legal systems. Existing retrieval methods often underperform in such domain-specific scenarios, necessitating tailored approaches. In this paper, we introduce a novel two-phase text retrieval pipeline optimized for Japanese legal datasets. Our method leverages advanced language models to achieve state-of-the-art performance, significantly improving retrieval efficiency and accuracy. To further enhance robustness and adaptability, we incorporate an ensemble model that integrates multiple retrieval strategies, resulting in superior outcomes across diverse tasks. Extensive experiments validate the effectiveness of our approach, demonstrating strong performance on both Japanese legal datasets and widely recognized benchmarks like MS-MARCO. Our work establishes new standards for text retrieval in domain-specific and general contexts, providing a comprehensive solution for addressing complex queries in legal and multilingual environments. △ Less

Submitted 26 December, 2024; originally announced December 2024.

Improving Pareto Set Learning for Expensive Multi-objective Optimization via Stein Variational Hypernetworks

Authors: Minh-Duc Nguyen, Phuong Mai Dinh, Quang-Huy Nguyen, Long P. Hoang, Dung D. Le

Abstract: Expensive multi-objective optimization problems (EMOPs) are common in real-world scenarios where evaluating objective functions is costly and involves extensive computations or physical experiments. Current Pareto set learning methods for such problems often rely on surrogate models like Gaussian processes to approximate the objective functions. These surrogate models can become fragmented, result… ▽ More Expensive multi-objective optimization problems (EMOPs) are common in real-world scenarios where evaluating objective functions is costly and involves extensive computations or physical experiments. Current Pareto set learning methods for such problems often rely on surrogate models like Gaussian processes to approximate the objective functions. These surrogate models can become fragmented, resulting in numerous small uncertain regions between explored solutions. When using acquisition functions such as the Lower Confidence Bound (LCB), these uncertain regions can turn into pseudo-local optima, complicating the search for globally optimal solutions. To address these challenges, we propose a novel approach called SVH-PSL, which integrates Stein Variational Gradient Descent (SVGD) with Hypernetworks for efficient Pareto set learning. Our method addresses the issues of fragmented surrogate models and pseudo-local optima by collectively moving particles in a manner that smooths out the solution space. The particles interact with each other through a kernel function, which helps maintain diversity and encourages the exploration of underexplored regions. This kernel-based interaction prevents particles from clustering around pseudo-local optima and promotes convergence towards globally optimal solutions. Our approach aims to establish robust relationships between trade-off reference vectors and their corresponding true Pareto solutions, overcoming the limitations of existing methods. Through extensive experiments across both synthetic and real-world MOO benchmarks, we demonstrate that SVH-PSL significantly improves the quality of the learned Pareto set, offering a promising solution for expensive multi-objective optimization problems. △ Less

Submitted 15 March, 2025; v1 submitted 23 December, 2024; originally announced December 2024.

Comments: Accepted to AAAI-25

Adaptive Two-Phase Finetuning LLMs for Japanese Legal Text Retrieval

Authors: Quang Hoang Trung, Nguyen Van Hoang Phuc, Le Trung Hoang, Quang Huu Hieu, Vo Nguyen Le Duy

Abstract: Text Retrieval (TR) involves finding and retrieving text-based content relevant to a user's query from a large repository, with applications in real-world scenarios such as legal document retrieval. While most existing studies focus on English, limited work addresses Japanese contexts. In this paper, we introduce a new dataset specifically designed for Japanese legal contexts and propose a novel t… ▽ More Text Retrieval (TR) involves finding and retrieving text-based content relevant to a user's query from a large repository, with applications in real-world scenarios such as legal document retrieval. While most existing studies focus on English, limited work addresses Japanese contexts. In this paper, we introduce a new dataset specifically designed for Japanese legal contexts and propose a novel two-phase pipeline tailored to this domain. In the first phase, the model learns a broad understanding of global contexts, enhancing its generalization and adaptability to diverse queries. In the second phase, the model is fine-tuned to address complex queries specific to legal scenarios. Extensive experiments are conducted to demonstrate the superior performance of our method, which outperforms existing baselines. Furthermore, our pipeline proves effective in English contexts, surpassing comparable baselines on the MS MARCO dataset. We have made our code publicly available on GitHub, and the model checkpoints are accessible via HuggingFace. △ Less

Submitted 3 December, 2024; originally announced December 2024.

SimSiam Naming Game: A Unified Approach for Emergent Communication and Representation Learning

Authors: Nguyen Le Hoang, Tadahiro Taniguchi, Tianwei Fang, Akira Taniguchi, Masatoshi Nagano

Abstract: Emergent Communication (EmCom) investigates how agents develop symbolic communication through interaction without predefined language. Recent frameworks, such as the Metropolis--Hastings Naming Game (MHNG), formulate EmCom as the learning of shared external representations negotiated through interaction under joint attention, without explicit success or reward feedback. However, MHNG relies on sam… ▽ More Emergent Communication (EmCom) investigates how agents develop symbolic communication through interaction without predefined language. Recent frameworks, such as the Metropolis--Hastings Naming Game (MHNG), formulate EmCom as the learning of shared external representations negotiated through interaction under joint attention, without explicit success or reward feedback. However, MHNG relies on sampling-based updates that suffer from high rejection rates in high-dimensional perceptual spaces, making the learning process sample-inefficient for complex visual datasets. In this work, we propose the SimSiam Naming Game (SSNG), a feedback-free EmCom framework that replaces sampling-based updates with a symmetric, self-supervised representation alignment objective between autonomous agents. Building on a variational inference--based probabilistic interpretation of self-supervised learning, SSNG formulates symbol emergence as an alignment process between agents' latent representations mediated by message exchange. To enable end-to-end gradient-based optimization, discrete symbolic messages are learned via a Gumbel--Softmax relaxation, preserving the discrete nature of communication while maintaining differentiability. Experiments on CIFAR-10 and ImageNet-100 show that the emergent messages learned by SSNG achieve substantially higher linear-probe classification accuracy than those produced by referential games, reconstruction games, and MHNG. These results indicate that self-supervised representation alignment provides an effective mechanism for feedback-free EmCom in multi-agent systems. △ Less

Submitted 14 June, 2026; v1 submitted 29 October, 2024; originally announced October 2024.

On Goodhart's law, with an application to value alignment

Authors: El-Mahdi El-Mhamdi, Lê-Nguyên Hoang

Abstract: ``When a measure becomes a target, it ceases to be a good measure'', this adage is known as {\it Goodhart's law}. In this paper, we investigate formally this law and prove that it critically depends on the tail distribution of the discrepancy between the true goal and the measure that is optimized. Discrepancies with long-tail distributions favor a Goodhart's law, that is, the optimization of the… ▽ More ``When a measure becomes a target, it ceases to be a good measure'', this adage is known as {\it Goodhart's law}. In this paper, we investigate formally this law and prove that it critically depends on the tail distribution of the discrepancy between the true goal and the measure that is optimized. Discrepancies with long-tail distributions favor a Goodhart's law, that is, the optimization of the measure can have a counter-productive effect on the goal. We provide a formal setting to assess Goodhart's law by studying the asymptotic behavior of the correlation between the goal and the measure, as the measure is optimized. Moreover, we introduce a distinction between a {\it weak} Goodhart's law, when over-optimizing the metric is useless for the true goal, and a {\it strong} Goodhart's law, when over-optimizing the metric is harmful for the true goal. A distinction which we prove to depend on the tail distribution. We stress the implications of this result to large-scale decision making and policies that are (and have to be) based on metrics, and propose numerous research directions to better assess the safety of such policies in general, and to the particularly concerning case where these policies are automated with algorithms. △ Less

Submitted 12 October, 2024; originally announced October 2024.

Comments: 47 pages, 11 figures

Enabling P-type Conduction in Bilayer WS2 with NbP Topological Semimetal Contacts

Authors: Lauren Hoang, Asir Intisar Khan, Robert K. A. Bennett, Hyun-mi Kim, Zhepeng Zhang, Marisa Hocking, Ae Rim Choi, Il-Kwon Oh, Andrew J. Mannix, Eric Pop

Abstract: Two-dimensional (2D) semiconductors are promising for low-power complementary metal oxide semiconductor (CMOS) electronics, which require ultrathin n- and p-type transistor channels. Among 2D semiconductors, WS2 is expected to have good conduction for both electrons and holes, but p-type WS2 transistors have been difficult to realize due to the relatively deep valence band and the presence of mid-… ▽ More Two-dimensional (2D) semiconductors are promising for low-power complementary metal oxide semiconductor (CMOS) electronics, which require ultrathin n- and p-type transistor channels. Among 2D semiconductors, WS2 is expected to have good conduction for both electrons and holes, but p-type WS2 transistors have been difficult to realize due to the relatively deep valence band and the presence of mid-gap states with conventional metal contacts. Here, we report topological semimetal NbP as p-type electrical contacts to bilayer WS2 with up to 5.8 microamperes per micron hole current at room temperature; this is the highest to date for sub 2 nm thin WS2 and more than 50 times larger than with metals like Ni or Pd. The p-type conduction is enabled by the simultaneously high work function and low density of states of the NbP, which reduce Fermi level pinning. These contacts are sputter-deposited at room temperature, an approach compatible with CMOS fabrication, a step towards enabling ultrathin WS2 semiconductors in future nanoelectronics. △ Less

Submitted 27 September, 2024; originally announced September 2024.

The poison of dimensionality

Authors: Lê-Nguyên Hoang

Abstract: This paper advances the understanding of how the size of a machine learning model affects its vulnerability to poisoning, despite state-of-the-art defenses. Given isotropic random honest feature vectors and the geometric median (or clipped mean) as the robust gradient aggregator rule, we essentially prove that, perhaps surprisingly, linear and logistic regressions with $D \geq 169 H^2/P^2$ paramet… ▽ More This paper advances the understanding of how the size of a machine learning model affects its vulnerability to poisoning, despite state-of-the-art defenses. Given isotropic random honest feature vectors and the geometric median (or clipped mean) as the robust gradient aggregator rule, we essentially prove that, perhaps surprisingly, linear and logistic regressions with $D \geq 169 H^2/P^2$ parameters are subject to arbitrary model manipulation by poisoners, where $H$ and $P$ are the numbers of honestly labeled and poisoned data points used for training. Our experiments go on exposing a fundamental tradeoff between augmenting model expressivity and increasing the poisoners' attack surface, on both synthetic data, and on MNIST & FashionMNIST data for linear classifiers with random features. We also discuss potential implications for source-based learning and neural nets. △ Less

Submitted 25 September, 2024; originally announced September 2024.

Comments: 29 pages, 3 figures

Non-ergodic inference for stationary-increment harmonizable stable processes

Authors: Ly Viet Hoang, Evgeny Spodarev

Abstract: We consider the class of stationary-increment harmonizable stable processes with infinite control measure, which most notably includes real harmonizable fractional stable motions. We give conditions for the integrability of the paths of such processes with respect to a finite, absolutely continuous measure and derive the distributional characteristics of the path integral with respect to said meas… ▽ More We consider the class of stationary-increment harmonizable stable processes with infinite control measure, which most notably includes real harmonizable fractional stable motions. We give conditions for the integrability of the paths of such processes with respect to a finite, absolutely continuous measure and derive the distributional characteristics of the path integral with respect to said measure. The convolution of the path of a stationary-increment harmonizable stable process with a suitable measure yields a real stationary harmonizable stable process with finite control measure. This allows us to construct consistent estimators for the index of stability as well as the kernel function in the integral representation of a stationary increment harmonizable stable process (up to a constant factor). For real harmonizable fractional stable motions consistent estimators for the index of stability and its Hurst parameter are given. These are computed directly from the periodogram frequency estimates of the smoothed process. △ Less

Submitted 19 August, 2024; originally announced August 2024.

Excitation laser energy dependence of the gap-mode TERS spectra of WS$_2$ and MoS$_2$ on silver

Authors: Andrey Krayev, Eleonora Isotta, Lauren Hoang, Jerry A. Yang, Kathryn Neilson, Minyuan Wang, Noah Haughn, Eric Pop, Andrew Mannix, Oluwaseyi Balogun, Chih-Feng Wang

Abstract: We present a systematic study of the dependence of gap mode tip-enhanced Raman scattering (TERS) of mono- and bi-layer WS$_2$ and MoS$_2$ as a function of excitation laser energy. We collected consecutive TERS maps of mono-and bi-layer regions with 6 different excitation lasers. To decrease the acquisition time, we used for the first time concurrent excitation and collection with two lasers simult… ▽ More We present a systematic study of the dependence of gap mode tip-enhanced Raman scattering (TERS) of mono- and bi-layer WS$_2$ and MoS$_2$ as a function of excitation laser energy. We collected consecutive TERS maps of mono-and bi-layer regions with 6 different excitation lasers. To decrease the acquisition time, we used for the first time concurrent excitation and collection with two lasers simultaneously. We found that the E$_{2g}$/A$_{1g}$ peak intensity ratio for bilayer WS$_2$@Ag and the A'/A$_{1g}$ peak intensity ratio of the out-of-plane modes for mono- and bilayer change in a significantly non-monotonous way with excitation laser energies from 1.58 to 2.62 eV. The former ratio increases at energies corresponding to A and B excitons in bilayer WS$_2$. The intensity of the A peak in the monolayer, and hence the A/A$_{1g}$ ratio, is surprisingly high at low excitation energies, dips dramatically at energy corresponding to the A exciton, and is restored partially in between A and B excitons, though still showing a descending trend with increasing energy. A similar picture was observed in mono- and bi-layer MoS$_2$, though the existing set of lasers did not match its excitonic profile as nicely as for WS$_2$. We attribute the observed behavior to intermediate (Fano resonance) or strong (Rabi splitting) coupling between the excitons in transition metal dichalcogenides (TMDs) and the plasmons in the tip-substrate nanocavity. This is akin to the so-called Fano (Rabi) transparency experimentally observed in far field scattering from TMDs between two plasmonic metals. The possibility of intermediate/strong coupling between excitonic resonances in TMDs and the nanocavity re-evaluates the role of resonances in gap-mode TERS and should become an important factor to be considered by TERS practitioners when planning experiments. Finally, we propose the ideal substrate for efficient TERS and tip enhanced photoluminescence measurements. △ Less

Submitted 18 July, 2024; originally announced July 2024.

Comments: 21 pages, 10 figures

The infrastructure powering IBM's Gen AI model development

Authors: Talia Gershon, Seetharami Seelam, Brian Belgodere, Milton Bonilla, Lan Hoang, Danny Barnett, I-Hsin Chung, Apoorve Mohan, Ming-Hung Chen, Lixiang Luo, Robert Walkup, Constantinos Evangelinos, Shweta Salaria, Marc Dombrowa, Yoonho Park, Apo Kayi, Liran Schour, Alim Alim, Ali Sydney, Pavlos Maniotis, Laurent Schares, Bernard Metzler, Bengi Karacali-Akyamac, Sophia Wen, Tatsuhiro Chiba , et al. (122 additional authors not shown)

Abstract: AI Infrastructure plays a key role in the speed and cost-competitiveness of developing and deploying advanced AI models. The current demand for powerful AI infrastructure for model training is driven by the emergence of generative AI and foundational models, where on occasion thousands of GPUs must cooperate on a single training job for the model to be trained in a reasonable time. Delivering effi… ▽ More AI Infrastructure plays a key role in the speed and cost-competitiveness of developing and deploying advanced AI models. The current demand for powerful AI infrastructure for model training is driven by the emergence of generative AI and foundational models, where on occasion thousands of GPUs must cooperate on a single training job for the model to be trained in a reasonable time. Delivering efficient and high-performing AI training requires an end-to-end solution that combines hardware, software and holistic telemetry to cater for multiple types of AI workloads. In this report, we describe IBM's hybrid cloud infrastructure that powers our generative AI model development. This infrastructure includes (1) Vela: an AI-optimized supercomputing capability directly integrated into the IBM Cloud, delivering scalable, dynamic, multi-tenant and geographically distributed infrastructure for large-scale model training and other AI workflow steps and (2) Blue Vela: a large-scale, purpose-built, on-premises hosting environment that is optimized to support our largest and most ambitious AI model training tasks. Vela provides IBM with the dual benefit of high performance for internal use along with the flexibility to adapt to an evolving commercial landscape. Blue Vela provides us with the benefits of rapid development of our largest and most ambitious models, as well as future-proofing against the evolving model landscape in the industry. Taken together, they provide IBM with the ability to rapidly innovate in the development of both AI models and commercial offerings. △ Less

Submitted 13 January, 2025; v1 submitted 7 July, 2024; originally announced July 2024.

Comments: Corresponding Authors: Talia Gershon, Seetharami Seelam,Brian Belgodere, Milton Bonilla

Mobility and Threshold Voltage Extraction in Transistors with Gate-Voltage-Dependent Contact Resistance

Authors: Robert K. A. Bennett, Lauren Hoang, Connor Cremers, Andrew J. Mannix, Eric Pop

Abstract: The mobility of emerging (e.g., two-dimensional, oxide, organic) semiconductors is commonly estimated from transistor current-voltage measurements. However, such devices often experience contact gating, i.e., electric fields from the gate modulate the contact resistance during measurements, which can lead conventional extraction techniques to estimate mobility incorrectly even by a factor >2. This… ▽ More The mobility of emerging (e.g., two-dimensional, oxide, organic) semiconductors is commonly estimated from transistor current-voltage measurements. However, such devices often experience contact gating, i.e., electric fields from the gate modulate the contact resistance during measurements, which can lead conventional extraction techniques to estimate mobility incorrectly even by a factor >2. This error can be minimized by measuring transistors at high gate-source bias, |$V_\mathrm{gs}$|, but this regime is often inaccessible in emerging devices that suffer from high contact resistance or early gate dielectric breakdown. Here, we propose a method of extracting mobility in transistors with gate-dependent contact resistance that does not require operation at high |$V_\mathrm{gs}$|, enabling accurate mobility extraction even in emerging transistors with strong contact gating. Our approach relies on updating the transfer length method (TLM) and can achieve <10% error even in regimes where conventional techniques overestimate mobility by >2$\times$. △ Less

Submitted 20 February, 2025; v1 submitted 29 April, 2024; originally announced April 2024.

Comments: Updated to include citation information for published manuscript on page 1; corrected direction of increasing channel length in Fig. 2b; minor changes to wording/phrasing throughout

The Heisenberg-RIXS instrument at the European XFEL

Authors: Justine Schlappa, Giacomo Ghiringhelli, Benjamin E. Van Kuiken, Martin Teichmann, Piter S. Miedema, Jan Torben Delitz, Natalia Gerasimova, Serguei Molodtsov, Luigi Adriano, Bernard Baranasic, Carsten Broers, Robert Carley, Patrick Gessler, Nahid Ghodrati, David Hickin, Le Phuong Hoang, Manuel Izquierdo, Laurent Mercadier, Giuseppe Mercurio, Sergii Parchenko, Marijan Stupar, Zhong Yin, Leonardo Martinelli, Giacomo Merzoni, Ying Ying Peng , et al. (22 additional authors not shown)

Abstract: Resonant Inelastic X-ray Scattering (RIXS) is an ideal X-ray spectroscopy method to push the combination of energy and time resolutions to the Fourier transform ultimate limit, because it is unaffected by the core-hole lifetime energy broadening. And in pump-probe experiments the interaction time is made very short by the same core-hole lifetime. RIXS is very photon hungry so it takes great advant… ▽ More Resonant Inelastic X-ray Scattering (RIXS) is an ideal X-ray spectroscopy method to push the combination of energy and time resolutions to the Fourier transform ultimate limit, because it is unaffected by the core-hole lifetime energy broadening. And in pump-probe experiments the interaction time is made very short by the same core-hole lifetime. RIXS is very photon hungry so it takes great advantage from high repetition rate pulsed X-ray sources like the European XFEL. The hRIXS instrument is designed for RIXS experiments in the soft X-ray range with energy resolution approaching the Fourier and the Heisenberg limits. It is based on a spherical grating with variable line spacing (VLS) and a position-sensitive 2D detector. Initially, two gratings are installed to adequately cover the whole photon energy range. With optimized spot size on the sample and small pixel detector the energy resolution can be better than 40 meV at any photon energy below 1000 eV. At the SCS instrument of the European XFEL the spectrometer can be easily positioned thanks to air-pads on a high-quality floor, allowing the scattering angle to be continuously adjusted over the 65-145 deg range. It can be coupled to two different sample interaction chamber, one for liquid jets and one for solids, each equipped at the state-of-the-art and compatible for optical laser pumping in collinear geometry. The measured performances, in terms of energy resolution and count rate on the detector, closely match design expectations. hRIXS is open to public users since the summer of 2022. △ Less

Submitted 13 March, 2024; originally announced March 2024.

Comments: 43 pages, 12 figures, Supplemental Material

Chemically Tailored Growth of 2D Semiconductors via Hybrid Metal-Organic Chemical Vapor Deposition

Authors: Zhepeng Zhang, Lauren Hoang, Marisa Hocking, Jenny Hu, Gregory Zaborski Jr., Pooja Reddy, Johnny Dollard, David Goldhaber-Gordon, Tony F. Heinz, Eric Pop, Andrew J. Mannix

Abstract: Two-dimensional (2D) semiconducting transition-metal dichalcogenides (TMDCs) are an exciting platform for new excitonic physics and next-generation electronics, creating a strong demand to understand their growth, doping, and heterostructures. Despite significant progress in solid-source (SS-) and metal-organic chemical vapor deposition (MOCVD), further optimization is necessary to grow highly cry… ▽ More Two-dimensional (2D) semiconducting transition-metal dichalcogenides (TMDCs) are an exciting platform for new excitonic physics and next-generation electronics, creating a strong demand to understand their growth, doping, and heterostructures. Despite significant progress in solid-source (SS-) and metal-organic chemical vapor deposition (MOCVD), further optimization is necessary to grow highly crystalline 2D TMDCs with controlled doping. Here, we report a hybrid MOCVD growth method that combines liquid-phase metal precursor deposition and vapor-phase organo-chalcogen delivery to leverage the advantages of both MOCVD and SS-CVD. Using our hybrid approach, we demonstrate WS$_2$ growth with tunable morphologies - from separated single-crystal domains to continuous monolayer films - on a variety of substrates, including sapphire, SiO$_2$, and Au. These WS$_2$ films exhibit narrow neutral exciton photoluminescence linewidths down to 33 meV and room-temperature mobility up to 34 - 36 cm$^2$V$^-$$^1$s$^-$$^1$). Through simple modifications to the liquid precursor composition, we demonstrate the growth of V-doped WS$_2$, MoxW$_1$$_-$$_x$S$_2$ alloys, and in-plane WS$_2$-MoS$_2$ heterostructures. This work presents an efficient approach for addressing a variety of TMDC synthesis needs on a laboratory scale. △ Less

Submitted 6 March, 2024; originally announced March 2024.