Searcharxiv⌕ Search

arXiv subjects

Hong Wang

Publications and source records attributed to Hong Wang.

At least 91 records · Page 5Linked to original sources

From Uniform to Adaptive: General Skip-Block Mechanisms for Efficient PDE Neural Operators

In recent years, Neural Operators(NO) have gradually emerged as a popular approach for solving Partial Differential Equations (PDEs). However, their application to large-scale engineering tasks suffers from significant computational overhead. And the fact that current models impose a uniform computational cost while physical fields exhibit vastly different complexities constitutes a fundamental mismatch, which is the root of this inefficiency. For instance, in turbulence flows, intricate vortex regions require deeper network processing compared to stable flows. To address this, we introduce a framework: Skip-Block Routing (SBR), a general framework designed for Transformer-based neural operators, capable of being integrated into their multi-layer architectures. First, SBR uses a routing mechanism to learn the complexity and ranking of tokens, which is then applied during inference. Then, in later layers, it decides how many tokens are passed forward based on this ranking. This way, the model focuses more processing capacity on the tokens that are more complex. Experiments demonstrate that SBR is a general framework that seamlessly integrates into various neural operators. Our method reduces computational cost by approximately 50% in terms of Floating Point Operations (FLOPs), while still delivering up to 2x faster inference without sacrificing accuracy.

cs.LG↗

Mixture-of-Experts Operator Transformer for Large-Scale PDE Pre-Training

Pre-training has proven effective in addressing data scarcity and performance limitations in solving PDE problems with neural operators. However, challenges remain due to the heterogeneity of PDE datasets in equation types, which leads to high errors in mixed training. Additionally, dense pre-training models that scale parameters by increasing network width or depth incur significant inference costs. To tackle these challenges, we propose a novel Mixture-of-Experts Pre-training Operator Transformer (MoE-POT), a sparse-activated architecture that scales parameters efficiently while controlling inference costs. Specifically, our model adopts a layer-wise router-gating network to dynamically select 4 routed experts from 16 expert networks during inference, enabling the model to focus on equation-specific features. Meanwhile, we also integrate 2 shared experts, aiming to capture common properties of PDE and reduce redundancy among routed experts. The final output is computed as the weighted average of the results from all activated experts. We pre-train models with parameters from 30M to 0.5B on 6 public PDE datasets. Our model with 90M activated parameters achieves up to a 40% reduction in zero-shot error compared with existing models with 120M activated parameters. Additionally, we conduct interpretability analysis, showing that dataset types can be inferred from router-gating network decisions, which validates the rationality and effectiveness of the MoE architecture.

cs.LG↗

Heilbronn's triangle problem in three dimensions

We show that among any $n$ points in the unit cube one can find a triangle of area at most $n^{-2/3-c}$ for some absolute constant $c >0$. This gives the first non-trivial upper bound for the three-dimensional version of Heilbronn's triangle problem. This estimate is a consequence of the following result about configurations of point-line pairs in $\mathbb R^3$: for $n \ge 2$ let $p_1, \ldots,p_n \in [0,1]^3$ be a collection of points and let $\ell_i$ be a line through $p_i$ for every $i$ such that $d(p_i, \ell_j) \ge δ$ for all $i\neq j$. Then we have $n \lesssim δ^{-3+γ}$ for some absolute constant $γ>0$. The analogous result about point-line configurations in the plane was previously established by Cohen, Pohoata and the last author.

math.CO↗

STNet: Spectral Transformation Network for Solving Operator Eigenvalue Problem

Operator eigenvalue problems play a critical role in various scientific fields and engineering applications, yet numerical methods are hindered by the curse of dimensionality. Recent deep learning methods provide an efficient approach to address this challenge by iteratively updating neural networks. These methods' performance relies heavily on the spectral distribution of the given operator: larger gaps between the operator's eigenvalues will improve precision, thus tailored spectral transformations that leverage the spectral distribution can enhance their performance. Based on this observation, we propose the Spectral Transformation Network (STNet). During each iteration, STNet uses approximate eigenvalues and eigenfunctions to perform spectral transformations on the original operator, turning it into an equivalent but easier problem. Specifically, we employ deflation projection to exclude the subspace corresponding to already solved eigenfunctions, thereby reducing the search space and avoiding converging to existing eigenfunctions. Additionally, our filter transform magnifies eigenvalues in the desired region and suppresses those outside, further improving performance. Extensive experiments demonstrate that STNet consistently outperforms existing learning-based methods, achieving state-of-the-art performance in accuracy.

cs.LG↗

Accelerating IC Thermal Simulation Data Generation via Block Krylov and Operator Action

Recent advances in data-driven approaches, such as neural operators (NOs), have shown substantial efficacy in reducing the solution time for integrated circuit (IC) thermal simulations. However, a limitation of these approaches is requiring a large amount of high-fidelity training data, such as chip parameters and temperature distributions, thereby incurring significant computational costs. To address this challenge, we propose a novel algorithm for the generation of IC thermal simulation data, named block Krylov and operator action (BlocKOA), which simultaneously accelerates the data generation process and enhances the precision of generated data. BlocKOA is specifically designed for IC applications. Initially, we use the block Krylov algorithm based on the structure of the heat equation to quickly obtain a few basic solutions. Then we combine them to get numerous temperature distributions that satisfy the physical constraints. Finally, we apply heat operators on these functions to determine the heat source distributions, efficiently generating precise data points. Theoretical analysis shows that the time complexity of BlocKOA is one order lower than the existing method. Experimental results further validate its efficiency, showing that BlocKOA achieves a 420-fold speedup in generating thermal simulation data for 5000 chips with varying physical parameters and IC structures. Even with just 4% of the generation time, data-driven approaches trained on the data generated by BlocKOA exhibits comparable performance to that using the existing method.

cs.AI↗

A Simple and Better Baseline for Visual Grounding

Visual grounding aims to predict the locations of target objects specified by textual descriptions. For this task with linguistic and visual modalities, there is a latest research line that focuses on only selecting the linguistic-relevant visual regions for object localization to reduce the computational overhead. Albeit achieving impressive performance, it is iteratively performed on different image scales, and at every iteration, linguistic features and visual features need to be stored in a cache, incurring extra overhead. To facilitate the implementation, in this paper, we propose a feature selection-based simple yet effective baseline for visual grounding, called FSVG. Specifically, we directly encapsulate the linguistic and visual modalities into an overall network architecture without complicated iterative procedures, and utilize the language in parallel as guidance to facilitate the interaction between linguistic modal and visual modal for extracting effective visual features. Furthermore, to reduce the computational cost, during the visual feature learning, we introduce a similarity-based feature selection mechanism to only exploit language-related visual features for faster prediction. Extensive experiments conducted on several benchmark datasets comprehensively substantiate that the proposed FSVG achieves a better balance between accuracy and efficiency beyond the current state-of-the-art methods. Code is available at https://github.com/jcwang0602/FSVG.

cs.CV↗

Sticky Kakeya sets and the sticky Kakeya conjecture

A Kakeya set is a compact subset of $\mathbb{R}^n$ that contains a unit line segment pointing in every direction. The Kakeya conjecture asserts that such sets must have Hausdorff and Minkowski dimension $n$. There is a special class of Kakeya sets, called sticky Kakeya sets. Sticky Kakeya sets exhibit an approximate multi-scale self-similarity, and sets of this type played an important role in Katz, Łaba, and Tao's groundbreaking 1999 work on the Kakeya problem. We propose a special case of the Kakeya conjecture, which asserts that sticky Kakeya sets must have Hausdorff and Minkowski dimension $n$. We prove this conjecture in three dimensions.

math.CA↗

Towards Foundation Models for Cryo-ET Subtomogram Analysis

Cryo-electron tomography (cryo-ET) enables in situ visualization of macromolecular structures, where subtomogram analysis tasks such as classification, alignment, and averaging are critical for structural determination. However, effective analysis is hindered by scarce annotations, severe noise, and poor generalization. To address these challenges, we take the first step towards foundation models for cryo-ET subtomograms. First, we introduce CryoEngine, a large-scale synthetic data generator that produces over 904k subtomograms from 452 particle classes for pretraining. Second, we design an Adaptive Phase Tokenization-enhanced Vision Transformer (APT-ViT), which incorporates adaptive phase tokenization as an equivariance-enhancing module that improves robustness to both geometric and semantic variations. Third, we introduce a Noise-Resilient Contrastive Learning (NRCL) strategy to stabilize representation learning under severe noise conditions. Evaluations across 24 synthetic and real datasets demonstrate state-of-the-art (SOTA) performance on all three major subtomogram tasks and strong generalization to unseen datasets, advancing scalable and robust subtomogram analysis in cryo-ET.

cs.CV↗

Can We Trust the AI Pair Programmer? Copilot for API Misuse Detection and Correction

API misuse introduces security vulnerabilities, system failures, and increases maintenance costs, all of which remain critical challenges in software development. Existing detection approaches rely on static analysis or machine learning-based tools that operate post-development, which delays defect resolution. Delayed defect resolution can significantly increase the cost and complexity of maintenance and negatively impact software reliability and user trust. AI-powered code assistants, such as GitHub Copilot, offer the potential for real-time API misuse detection within development environments. This study evaluates GitHub Copilot's effectiveness in identifying and correcting API misuse using MUBench, which provides a curated benchmark of misuse cases. We construct 740 misuse examples, manually and via AI-assisted variants, using correct usage patterns and misuse specifications. These examples and 147 correct usage cases are analyzed using Copilot integrated in Visual Studio Code. Copilot achieved a detection accuracy of 86.2%, precision of 91.2%, and recall of 92.4%. It performed strongly on common misuse types (e.g., missing-call, null-check) but struggled with compound or context-sensitive cases. Notably, Copilot successfully fixed over 95% of the misuses it identified. These findings highlight both the strengths and limitations of AI-driven coding assistants, positioning Copilot as a promising tool for real-time pair programming and detecting and fixing API misuses during software development.

cs.SE↗

DriveSOTIF: Advancing Perception SOTIF Through Multimodal Large Language Models

Human drivers possess spatial and causal intelligence, enabling them to perceive driving scenarios, anticipate hazards, and react to dynamic environments. In contrast, autonomous vehicles lack these abilities, making it challenging to manage perception-related Safety of the Intended Functionality (SOTIF) risks, especially under complex or unpredictable driving conditions. To address this gap, we propose fine-tuning multimodal large language models (MLLMs) on a customized dataset specifically designed to capture perception-related SOTIF scenarios. Benchmarking results show that fine-tuned MLLMs achieve an 11.8\% improvement in close-ended VQA accuracy and a 12.0\% increase in open-ended VQA scores compared to baseline models, while maintaining real-time performance with a 0.59-second average inference time per image. We validate our approach through real-world case studies in Canada and China, where fine-tuned models correctly identify safety risks that challenge even experienced human drivers. This work represents the first application of domain-specific MLLM fine-tuning for SOTIF domain in autonomous driving. The dataset and related resources are available at github.com/s95huang/DriveSOTIF.git

cs.RO↗

Unveiling the critical factors in crystal structure graph representation: a comparative analysis using streamlined MLPSets frameworks

Graph Neural Networks have rapidly advanced in materials science and chemistry,with their performance critically dependent on comprehensive representations of crystal or molecular structures across five dimensions: elemental information, geometric topology, electronic interactions, symmetry, and long-range interactions. Existing models still exhibit limitations in representing electronic interactions, symmetry, and long-range information. This study compares physics-based site feature calculators with data-driven graph representation strategies. We find that the latter achieve superior performance in representation completeness, convergence speed, and extrapolation capability by incorporating electronic structure generation models-such as variational autoencoders (VAEs) that compress Kohn-Sham wave functions and leveraging multi-task learning. Notably, the CHGNet-V1/V2 strategies, when integrated into the DenseGNN model,significantly outperform state-of-the-art models across 35 datasets from Matbench and JARVIS-DFT, yielding predictions with accuracy close to that of DFT calculations. Furthermore, applying a pre-training and fine-tuning strategy substantially reduces the prediction error for band gaps of complex disordered materials, demonstrating the superiority and potential of data-driven graph representations in accelerating materials discovery.

cond-mat.mtrl-sci↗

Mechanical non-reciprocity programmed by shear jamming in soft composite solids

Mechanical non-reciprocity-manifested as asymmetric responses to opposing mechanical stimuli-has traditionally been achieved through intricate structural nonlinearities in metamaterials. However, continuum solids with inherent non-reciprocal mechanics remain underexplored, despite their promising potential for applications such as wave guiding, robotics, and adaptive materials. Here, we introduce a design principle by employing the shear jamming transition from granular physics to engineer non-reciprocal mechanics in soft composite solids. Through the control of the interplay between inclusion contact networks and matrix elasticity, we achieve tunable, direction-dependent asymmetry in both shear and normal mechanical responses. In addition to static regimes, we demonstrate programmable non-reciprocal dynamics by combining responsive magnetic profiles with the anisotropic characteristics of shear-jammed systems. This strategy enables asymmetric spatiotemporal control over motion transmission, a previously challenging feat in soft materials. Our work establishes a novel paradigm for designing non-reciprocal matter, bridging granular physics with soft material engineering to realize functionalities essential for mechano-intelligent systems.

cond-mat.soft↗

Unlocking the Potential of MLLMs in Referring Expression Segmentation via a Light-weight Mask Decoder

Reference Expression Segmentation (RES) aims to segment image regions specified by referring expressions and has become popular with the rise of multimodal large models (MLLMs). While MLLMs excel in semantic understanding, their token-generation paradigm struggles with pixel-level dense prediction. Existing RES methods either couple MLLMs with the parameter-heavy Segment Anything Model (SAM) with 632M network parameters or adopt SAM-free lightweight pipelines that sacrifice accuracy. To address the trade-off between performance and cost, we specifically propose MLLMSeg, a novel framework that fully exploits the inherent visual detail features encoded in the MLLM vision encoder without introducing an extra visual encoder. Besides, we propose a detail-enhanced and semantic-consistent feature fusion module (DSFF) that fully integrates the detail-related visual feature with the semantic-related feature output by the large language model (LLM) of MLLM. Finally, we establish a light-weight mask decoder with only 34M network parameters that optimally leverages detailed spatial features from the visual encoder and semantic features from the LLM to achieve precise mask prediction. Extensive experiments demonstrate that our method generally surpasses both SAM-based and SAM-free competitors, striking a better balance between performance and cost. Code is available at https://github.com/jcwang0602/MLLMSeg.

cs.CV↗

AdS$_{3}$ spacetime wormhole fluctuations and their impact on false vacuum decay

In this work, we studied the boundary wormhole fluctuations of AdS$_{3}$ spacetime and their impact on false vacuum decay. The model consists of three dimensional general relativity coupled to a real scalar field with self-interaction potential. Our analysis focuses on handlebody spacetimes, which correspond to boundary wormhole fluctuations. These fluctuations result in variations in the spacetime genus. Using the dilute wormhole gas approximation, we computed the boundary wormhole creation rate in the semiclassical region. Our results indicate that boundary wormholes are more likely to be created in spacetimes with larger absolute values of the cosmological constant. Furthermore, we find that smaller wormholes are more easily created compared to larger ones. Additionally, we investigated the impact of boundary wormhole fluctuations on false vacuum decay. The vacuum decay rate increases when the effects of wormhole fluctuations are considered. We show that all results remain qualitatively unchanged when certain parameter values are varied.

hep-th↗

Impacts of wormholes on false vacuum bubble tunneling

In this study, we generalize the work of Farhi, Guth and Guven [Nucl. Phys. B 339 (1990) 417] to include a wormhole effect. We study the influence of the wormhole on the tunneling of the false vacuum bubble. The spherically symmetric bubble has two domain walls. The classical dynamics of each domain wall are constrained by two classically forbidden regions. We find that the wormhole has increased the number of instantons and thus can enhance the tunneling rate. We have analytically derived the formula for the bubble tunneling rate, which is consistent with the result obtained by Farhi, Guth and Guven when the wormhole disappears. We show that the tunneling rate increases with the throat radius of the wormhole. Furthermore, we illustrate that the tunneling rate increases with an increases in the mass of the black hole or the surface tension. In the case where a wormhole exists within a bubble and is connected to another universe, we demonstrate that the bubble can enter into another universe through quantum tunneling.

gr-qc↗

Lower Bounds for Error Coefficients of Griesmer Optimal Linear Codes via Iteration

The error coefficient of a linear code is defined as the number of minimum-weight codewords. In an additive white Gaussian noise channel, optimal linear codes with the smallest error coefficients achieve the best possible asymptotic frame error rate (AFER) among all optimal linear codes under maximum likelihood decoding. Such codes are referred to as AFER-optimal linear codes. The Griesmer bound is essential for determining the optimality of linear codes. However, establishing tight lower bounds on the error coefficients of Griesmer optimal linear codes is challenging, and the linear programming bound often performs inadequately. In this paper, we propose several iterative lower bounds for the error coefficients of Griesmer optimal linear codes. Specifically, for binary linear codes, our bounds are tight in most cases when the dimension does not exceed $5$. To evaluate the performance of our bounds when they are not tight, we also determine the parameters of the remaining 5-dimensional AFER-optimal linear codes. Our final comparison demonstrates that even when our bounds are not tight, they remain very close to the actual values, with a gap of less than or equal to $2$.

cs.IT↗

"Who Should I Believe?": User Interpretation and Decision-Making When a Family Healthcare Robot Contradicts Human Memory

Advancements in robotic capabilities for providing physical assistance, psychological support, and daily health management are making the deployment of intelligent healthcare robots in home environments increasingly feasible in the near future. However, challenges arise when the information provided by these robots contradicts users' memory, raising concerns about user trust and decision-making. This paper presents a study that examines how varying a robot's level of transparency and sociability influences user interpretation, decision-making and perceived trust when faced with conflicting information from a robot. In a 2 x 2 between-subjects online study, 176 participants watched videos of a Furhat robot acting as a family healthcare assistant and suggesting a fictional user to take medication at a different time from that remembered by the user. Results indicate that robot transparency influenced users' interpretation of information discrepancies: with a low transparency robot, the most frequent assumption was that the user had not correctly remembered the time, while with the high transparency robot, participants were more likely to attribute the discrepancy to external factors, such as a partner or another household member modifying the robot's information. Additionally, participants exhibited a tendency toward overtrust, often prioritizing the robot's recommendations over the user's memory, even when suspecting system malfunctions or third-party interference. These findings highlight the impact of transparency mechanisms in robotic systems, the complexity and importance associated with system access control for multi-user robots deployed in home environments, and the potential risks of users' over reliance on robots in sensitive domains such as healthcare.

cs.HC↗

Sum Rate Maximization for Pinching Antennas Assisted RSMA System With Multiple Waveguides

In this letter, a pinching antennas (PAs) assisted rate splitting multiple access (RSMA) system with multiple waveguides is investigated to maximize sum rate. A two-step algorithm is proposed to determine PA activation scheme and optimize the waveguide beamforming. Specifically, a low complexity spatial correlation and distance based method is proposed for PA activation selection. After determining the PA activation status, a semi-definite programming (SDP) based successive convex approximation (SCA) is leveraged to obtain the optimal waveguide beamforming. Simulation results show that the proposed multiple waveguides based PAs assisted RSMA method achieves better performance than various benchmarking schemes.

eess.SP↗