arXiv subjects

Huan Wu

Publications and source records attributed to Huan Wu.

At least 19 recordsLinked to original sources

arXiv2026

SenseNova-U1.5: Towards Native Unified Visual Intelligence

We launch SenseNova-U1.5, an 8B-MoT native unified multimodal model that understands, reasons about, and generates visual content within an encoder-free and VAE-free architecture. We strengthen its visual interface through spatially coherent patch reconstruction and scale its training with carefully curated generation and editing data, improved task formulation, structural prompt enhancement, and native resolutions of up to 4K. For post-training, we optimize specialized experts for visual aesthetics, bilingual text rendering, infographic generation, and image editing, and consolidate their capabilities through multi-expert on-policy distillation. Across extensive evaluations, SenseNova-U1.5 largely advances image fidelity, text rendering, complex composition, multi-reference editing, and interleaved generation, while improving instruction following and preserving subject identity, geometry, and unmodified regions. Despite limited exposure to structured formats in its generation data, SenseNova-U1.5 generalizes effectively to long, complex, and structured visual instructions, further proving that multimodal understanding can transfer to visual planning and creation. Together, these findings position native unified modelling as a promising path towards systems that perceive, reason and create within a fully end-to-end framework. We will open-source training code, including supervised fine-tuning, reinforcement learning, and on-policy distillation.

cs.CV↗

arXiv2026

LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering

African American English (AAE), a rule-governed dialect spoken by over 30 million people, is routinely misinterpreted and "corrected" by large language models (LLMs). Across six instruction-tuned LLMs (14B to 70B), we show that state-of-the-art models systematically prefer Standard American English (SAE) continuations even when the preceding context is in AAE, effectively rewriting AAE into SAE. We present an end-to-end framework to audit and mitigate this bias. For auditing, we introduce conditional Dialect Group Invariance (cDGI), which isolates true model bias from translator-induced artifacts, and a feature-level localization analysis that identifies which AAE markers most strongly trigger bias; we find that syntactic constructions, especially negative concord (e.g., "ain't nobody"), are universal triggers across all models. For mitigation, we introduce, to our knowledge, the first application of activation steering to dialect bias: a training-free, test-time method that extracts dialect directions via causal tracing and injects them into bias-relevant layers. Activation steering reduces bias 5 to 20 times more than prompting while preserving SAE fluency. To enable this work, we release REAL-AAE , the largest real-AAE parallel corpus to date: 17,479 AAE/SAE/ AAE_back triplets from natural tweets (2 to 6 times larger than prior real-AAE resources), validated automatically (BERTScore F1 = 0.95) and by three native AAE speakers (83.0% semantic agreement).

cs.CL↗

arXiv2026

Nonperturbative isotope effect on light-matter interaction in boron arsenide

The interaction of light and matter under strong isotope disorder gives rise to unconventional physics that goes beyond the quantum perturbation theory. In boron arsenide, the large mass difference between the two stable boron isotopes presents a paradigmatic case where perturbation theory fails, yet a unified theoretical framework across the perturbative and nonperturbative regime has remained elusive. Here, we develop a nonperturbative approach to capture isotope-disorder effect in boron arsenide, which fundamentally alters light-matter interactions. We reveal that coherently mixed vibrations between two boron isotopes reshape the dielectric function in the nonperturbative regime. The nonperturbative isotope interactions dictate the properties of coupled surface phonon polaritons and near-field radiative heat transfer. Two-fold tuning of radiative heat flux is achieved by modulating the surface phonon polariton resonance via isotope engineering. This work establishes a unified framework connecting the perturbative and nonperturbative limits, enabling quantitative predictions across weak to strong disorder regime.

cond-mat.mtrl-sci↗

arXiv2026

Pressure and strain tuning of the alternating bilayer-trilayer Ruddlesden-Popper nickelate: crystal and electronic structure

We use first-principles calculations to investigate the crystal and electronic structure of the hybrid bilayer-trilayer Ruddlesden-Popper (RP) nickelate La$_7$Ni$_5$O$_{17}$ under hydrostatic pressure and biaxial compressive strain. By analyzing the irreducible representations of the dynamically unstable phonon modes in the high-symmetry $P4/mmm$ structure, we identify a dynamically stable lower-symmetry $C2/c$ structure containing octahedral tilts. The application of both pressure and compressive strain tends to suppress the octahedral tilts, effectively tetragonalizing the structure, in analogy with the conventional RPs. The electronic structure under hydrostatic pressure and strain has similarities, but it differs in the position of the $d_{z^2}$ bonding band from the trilayer block. This band crosses the Fermi level at a pressure of 30 GPa, but it remains below it for any level of compressive strain. This strain-induced modification mirrors the electronic structure changes observed in the conventional bilayer nickelate.

cond-mat.mtrl-sci↗

arXiv2026

DisenReason: Behavior Disentanglement and Latent Reasoning for Shared-Account Sequential Recommendation

Shared-account usage is common on streaming and e-commerce platforms, where multiple users share one account. Existing shared-account sequential recommendation (SSR) methods often assume a fixed number of latent users per account, limiting their ability to adapt to diverse sharing patterns and reducing recommendation accuracy. Recent latent reasoning technique applied in sequential recommendation (SR) generate intermediate embeddings from the user embedding (e.g, last item embedding) to uncover users' potential interests, which inspires us to treat the problem of inferring the number of latent users as generating a series of intermediate embeddings, shifting from inferring preferences behind user to inferring the users behind account. However, the last item cannot be directly used for reasoning in SSR, as it can only represent the behavior of the most recent latent user, rather than the collective behavior of the entire account. To address this, we propose DisenReason, a two-stage reasoning method tailored to SSR. DisenReason combines behavior disentanglement stage from frequency-domain perspective to create a collective and unified account behavior representation, which serves as a pivot for latent user reasoning stage to infer the number of users behind the account. Experiments on four benchmark datasets show that DisenReason consistently outperforms all state-of-the-art baselines across four benchmark datasets, achieving relative improvements of up to 12.56\% in MRR@5 and 6.06\% in Recall@20.

cs.IR↗

arXiv2026

AIDABench: AI Data Analytics Benchmark

As AI-driven document understanding and processing tools become increasingly prevalent in real-world applications, the need for rigorous evaluation standards has grown increasingly urgent. Existing benchmarks and evaluations often focus on isolated capabilities or simplified scenarios, failing to capture the end-to-end task effectiveness required in practical settings. To address this gap, we introduce AIDABench, a comprehensive benchmark for evaluating AI systems on complex data analytics tasks in an end-to-end manner. AIDABench encompasses 600+ diverse document analysis tasks across three core capability dimensions: question answering, data visualization, and file generation. These tasks are grounded in realistic scenarios involving heterogeneous data types, including spreadsheets, databases, financial reports, and operational records, and reflect analytical demands across diverse industries and job functions. Notably, the tasks in AIDABench are sufficiently challenging that even human experts require 1-2 hours per question when assisted by AI tools, underscoring the benchmark's difficulty and real-world complexity. We evaluate 11 state-of-the-art models on AIDABench, spanning both proprietary (e.g., Claude Sonnet 4.5, Gemini 3 Pro Preview) and open-source (e.g., Qwen3-Max-2026-01-23-Thinking) families. Our results reveal that complex, real-world data analytics tasks remain a significant challenge for current AI systems, with the best-performing model achieving only 59.43% pass-at-1. We provide a detailed analysis of failure modes across each capability dimension and identify key challenges for future research. AIDABench offers a principled reference for enterprise procurement, tool selection, and model optimization, and is publicly available at https://github.com/MichaelYang-lyx/AIDABench.

cs.AI↗

arXiv2026

Visible Light Positioning With Lam\'e Curve LEDs: A Generic Approach for Camera Pose Estimation

Camera-based visible light positioning (VLP) is a promising technique for accurate and low-cost indoor camera pose estimation (CPE). To reduce the number of required light-emitting diodes (LEDs), advanced methods commonly exploit LED shape features for positioning. Although interesting, they are typically restricted to a single LED geometry, leading to failure in heterogeneous LED-shape scenarios. To address this challenge, this paper investigates Lam\'e curves as a unified representation of common LED shapes and proposes a generic VLP algorithm using Lam\'e curve-shaped LEDs, termed LC-VLP. In the considered system, multiple ceiling-mounted Lam\'e curve-shaped LEDs periodically broadcast their curve parameters via visible light communication, which are captured by a camera-equipped receiver. Based on the received LED images and curve parameters, the receiver can estimate the camera pose using LC-VLP. Specifically, an LED database is constructed offline to store the curve parameters, while online positioning is formulated as a nonlinear least-squares problem and solved iteratively. To provide a reliable initialization, a correspondence-free perspective-n-points (FreePnP) algorithm is further developed, enabling approximate CPE without any pre-calibrated reference points. The performance of LC-VLP is verified by both simulations and experiments. Simulations show that LC-VLP outperforms state-of-the-art methods in both circular- and rectangular-LED scenarios. Compared to a perspective arcs algorithm, LC-VLP can achieve reductions of both over 30% in average position and rotation errors. Experiments further show that LC-VLP can achieve an average position accuracy of less than 4 cm.

eess.SP↗

arXiv2026

An altermagnetic materials library in intercalated transition-metal dichalcogenides

Altermagnets represent a promising class of magnetic materials owing to their distinctive spin-split band structures in the absence of net magnetization. Here, we present a first-principles investigation of altermagnetism in magnetically intercalated transition metal dichalcogenides (TMDs) with the general formula T$_y$MX$_2$ (T= 3$d$-transition metal, M= transition-metal, X=chalcogen, $y$= 1/3 or 1/4). For a TMD host with 2H structure, compounds exhibiting A-type antiferromagnetism are $g$-wave altermagnets by symmetry. We identify several intercalated TMDs fulfilling the conditions for altermagnetic order to be realized. Several of these candidate materials display spin-splittings at the Fermi level as large as 100 meV.

cond-mat.mtrl-sci↗

arXiv2025

Unobserved Heterogeneous Spillover Effects in Instrumental Variable Models

This paper develops a general framework for identifying causal effects in settings with spillovers, where both outcomes and endogenous treatment decisions are influenced by peers within a known group. It introduces the generalized local average controlled spillover and direct effects (LACSEs and LACDEs), which extend the local average treatment effect framework to settings with spillovers and establish sufficient conditions for their point identification without restricting the cardinality of the support of instrumental variables. These conditions clarify the necessity of commonly imposed restrictions to achieve point identification with binary instruments in related studies. The paper then defines the marginal controlled spillover and direct effects (MCSEs and MCDEs), which naturally extend the marginal treatment effect framework to settings with spillovers and are nonparametrically point identified from continuous variation in instruments. These marginal effects serve as building blocks for a broad class of policy-relevant treatment effects, including some causal spillover parameters in the related literature. Semiparametric and parametric estimators are developed, and an application using Add Health data reveals heterogeneity in education spillovers within best-friend networks.

econ.EM↗

arXiv2025

AES-SpMM: Balancing Accuracy and Speed by Adaptive Edge Sampling Strategy to Accelerate SpMM in GNNs

Coordinating the design of sampling and sparse-dense matrix multiplication (SpMM) is crucial for accelerating graph neural networks (GNNs). However, due to irrational sampling strategies, existing methods face a trade-off between accuracy and speed. Moreover, as computational optimizations progress, data loading has gradually become the primary bottleneck in GNN inference. To address these issues, we propose AES-SpMM, an adaptive edge sampling SpMM kernel. It considers the relationship between the number of non-zero elements in each matrix row and the shared memory width. The edge sampling scheme is adaptively selected according to the different situations of each row. AES-SpMM reduces the graph size through adaptive edge sampling to fit the GPU's shared memory, lowering the computational cost and enhancing data locality, thus balancing the accuracy and speed of GNN inference. Additionally, we introduce a quantization-based AES-SpMM, which applies quantization and dequantization to feature data in GNNs. This approach significantly reduces data loading time while keeping accuracy loss negligible. We evaluated AES-SpMM with common GNN models and datasets. The results show that AES-SpMM outperforms both the cuSPARSE SpMM kernel and GE-SpMM by up to 25.87 times and 23.01 times, respectively, with less than 1% accuracy loss. Compared to ES-SpMM, it reduces accuracy loss by 3.4% on average , achieving a 1.31 times speedup. Compared to AES-SpMM, quantization-based AES-SpMM has a maximum accuracy loss of 0.3% and feature data loading time overhead is reduced by 50.91%-70.51%.

cs.DC↗

arXiv2025

Robust Identification in Randomized Experiments with Noncompliance

Instrument variable (IV) methods are widely used in empirical research to identify causal effects of a policy. In the local average treatment effect (LATE) framework, the IV estimand identifies the LATE under three main assumptions: random assignment, exclusion restriction, and monotonicity. However, these assumptions are often questionable in many applications, leading some researchers to doubt the causal interpretation of the IV estimand. This paper considers a robust identification of causal parameters in a randomized experiment setting with noncompliance where the standard LATE assumptions could be violated. We discuss identification under two sets of weaker assumptions: random assignment and exclusion restriction (without monotonicity), and random assignment and monotonicity (without exclusion restriction). We derive sharp bounds on some causal parameters under these two sets of relaxed LATE assumptions. Finally, we apply our method to revisit the random information experiment conducted in Bursztyn, González, and Yanagizawa-Drott (2020) and find that the standard LATE assumptions are jointly incompatible in this application. We then estimate the robust identified sets under the two sets of relaxed assumptions.

econ.EM↗

arXiv2025

DCI: A Coordinated Allocation and Filling Workload-Aware Dual-Cache Allocation GNN Inference Acceleration System

Graph Neural Networks (GNNs) are powerful tools for processing graph-structured data, increasingly used for large-scale real-world graphs via sampling-based inference methods. However, inherent characteristics of neighbor sampling lead to redundant data loading during GNN inference, compounded by inefficient data transfers between host and GPU memory, resulting in slow inference and low resource utilization. Existing methods to accelerate GNN inference face several challenges: (1) low practical GPU memory utilization, (2) overlooking adjacency matrix locality, and (3) long preprocessing time. To address these challenges, we introduce DCI, an efficient workload-aware dual-cache allocation system for GNN inference acceleration. DCI allocates cache capacities for both node features and adjacency matrices based on workload patterns during the pre-sampling phase, leveraging a lightweight cache-filling algorithm to optimize data loading efficiency. Experimental results demonstrate that DCI accelerates sampling and node feature loading, achieving end-to-end inference speedups of 1.18$\times$ to 11.26$\times$ compared to DGL, and 1.14$\times$ to 13.68$\times$ over RAIN, while reducing preprocessing time by 52.8\% to 98.7\%. Additionally, DCI outperforms state-of-the-art single-cache inference systems by achieving speedup of 1.08$\times$ to 1.32$\times$. We also compared DCI with DUCATI's dual-cache population strategy. Our lightweight population algorithm allows DCI to achieve nearly the same inference speed while keeping preprocessing time to less than 20\% of that required by DUCATI.

cs.AR↗

arXiv2025

The existence and stability of viscosity solutions to perturbed contact Hamilton-Jacobi equations

We consider a contact Hamiltonian $H(x,p,u)$ with certain dependence on the contact variable $u$. If $u_{-}$ is a viscosity solution of the contact Hamilton-Jacobi equation \[H(x,D_{x}u(x),u(x))=0,\quad x\in M,\] and $u_{-}$ is locally Lyapunov asymptotically stable, we will prove that the perturbed equation \[H(x,D_{x}u(x),u(x))+\varepsilon P(x,D_{x}u(x),u(x))=0,\quad x\in M,\] does exist viscosity solution $u_{-}^{\varepsilon}$ which converges uniformly to $u_{-}$, as perturbation parameter $\varepsilon$ converges to 0. Moreover, we give a case that in a neighborhood of viscosity solution $u_-$, the perturbed equation has an unique viscosity solution $u_{-}^{\varepsilon}$. Furthermore, $u_{-}^{\varepsilon}$ keeps locally Lyapunov asymptotically stability.

math.AP↗

arXiv2024

Testing Identifying Assumptions in Parametric Separable Models: A Conditional Moment Inequality Approach

In this paper, we propose a simple method for testing identifying assumptions in parametric separable models, namely treatment exogeneity, instrument validity, and/or homoskedasticity. We show that the testable implications can be written in the intersection bounds framework, which is easy to implement using the inference method proposed in Chernozhukov, Lee, and Rosen (2013), and the Stata package of Chernozhukov et al. (2015). Monte Carlo simulations confirm that our test is consistent and controls size. We use our proposed method to test the validity of some commonly used instrumental variables, such as the average price in other markets in Nevo and Rosen (2012), the Bartik instrument in Card (2009), and the test rejects both instrumental variable models. When the identifying assumptions are rejected, we discuss solutions that allow researchers to identify some causal parameters of interest after relaxing functional form assumptions. We show that the IV model is nontestable if no functional form assumption is made on the outcome equation, when there exists a one-to-one mapping between the continuous treatment variable, the instrument, and the first-stage unobserved heterogeneity.

econ.EM↗

arXiv2024

First Principles Validation of Energy Barriers in Ni$_{75}$Al$_{25}$

Precipitates in Nickel-based superalloys form during heat treatment on a time scale inaccessible to direct molecular dynamics simulation, but could be studied using kinetic Monte Carlo (KMC). This requires reliable values for the barrier energies separating distinct configurations over the trajectory of the system. In this study, we validate vacancy migration barriers found with the Activation-Relaxation Technique nouveau (ARTn) method in partially ordered Ni$_{75}$Al$_{25}$ with a monovacancy using published potentials for the atomic interactions against first-principles methods. In a first step, we confirm that the ARTn barrier energies agree with those determined with the nudged elastic band (NEB) method. As the number of atoms used in those calculations is too great for direct ab initio calculations, we then cut the cell size to 255 atoms, thus controlling finite size effects. We then use the plane-wave density functional theory (DFT) code CASTEP and its inbuilt NEB method in the smaller cells. This provides us with a continuous validation chain from first principles to kinetic Monte Carlo simulations with interatomic potentials. We then evaluate the barrier energies of five further interatomic potentials with NEB, demonstrating that none yields these with sufficient reliability for KMC simulations, with some of them failing completely. This is a first step towards quantifying the errors incurred in KMC simulations of precipitate formation and evolution.

cond-mat.mtrl-sci↗

arXiv2023

Smart Fuzzing of 5G Wireless Software Implementation

In this paper, we introduce a comprehensive approach to bolstering the security, reliability, and comprehensibility of OpenAirInterface5G (OAI5G), an open-source software framework for the exploration, development, and testing of 5G wireless communication systems. Firstly, we employ AFL++, a powerful fuzzing tool, to fuzzy-test OAI5G with respect to its configuration files rigorously. This extensive testing process helps identify errors, defects, and security vulnerabilities that may evade conventional testing methods. Secondly, we harness the capabilities of Large Language Models such as Google Bard to automatically decipher and document the meanings of parameters within the OAI5G codebase that are used in fuzzing. This automated parameter interpretation streamlines subsequent analyses and facilitates more informed decision-making. Together, these two techniques contribute to fortifying the OAI5G system, making it more robust, secure, and understandable for developers and analysts alike.

cs.SE↗

arXiv2023

Leveraging Optical Communication Fiber and AI for Distributed Water Pipe Leak Detection

Detecting leaks in water networks is a costly challenge. This article introduces a practical solution: the integration of optical network with water networks for efficient leak detection. Our approach uses a fiber-optic cable to measure vibrations, enabling accurate leak identification and localization by an intelligent algorithm. We also propose a method to access leak severity for prioritized repairs. Our solution detects even small leaks with flow rates as low as 0.027 L/s. It offers a cost-effective way to improve leak detection, enhance water management, and increase operational efficiency.

eess.SY↗

arXiv2022

Compressed domain vibration detection and classification for distributed acoustic sensing

Distributed acoustic sensing (DAS) is a novel enabling technology that can turn existing fibre optic networks to distributed acoustic sensors. However, it faces the challenges of transmitting, storing, and processing massive streams of data which are orders of magnitude larger than that collected from point sensors. The gap between intensive data generated by DAS and modern computing system with limited reading/writing speed and storage capacity imposes restrictions on many applications. Compressive sensing (CS) is a revolutionary signal acquisition method that allows a signal to be acquired and reconstructed with significantly fewer samples than that required by Nyquist-Shannon theorem. Though the data size is greatly reduced in the sampling stage, the reconstruction of the compressed data is however time and computation consuming. To address this challenge, we propose to map the feature extractor from Nyquist-domain to compressed-domain and therefore vibration detection and classification can be directly implemented in compressed-domain. The measured results show that our framework can be used to reduce the transmitted data size by 70% while achieves 99.4% true positive rate (TPR) and 0.04% false positive rate (TPR) along 5 km sensing fibre and 95.05% classification accuracy on a 5-class classification task.

eess.SP↗