Yuting Hu

Yuting Hu, Dancheng Liu, Qingyun Wang et al.

Identifying effective interventions from the scientific literature is challenging due to the high volume of publications, specialized terminology, and inconsistent reporting formats, making manual curation laborious and prone to oversight. To address this challenge, this paper proposes a novel framework leveraging large language models (LLMs), which integrates a progressive ontology prompting (POP) algorithm with a dual-agent system, named LLM-Duo. On the one hand, the POP algorithm conducts a prioritized breadth-first search (BFS) across a predefined ontology, generating structured prompt templates and action sequences to guide the automatic annotation process. On the other hand, the LLM-Duo system features two specialized LLM agents, an explorer and an evaluator, working collaboratively and adversarially to continuously refine annotation quality. We showcase the real-world applicability of our framework through a case study focused on speech-language intervention discovery. Experimental results show that our approach surpasses advanced baselines, achieving more accurate and comprehensive annotations through a fully automated process. Our approach successfully identified 2,421 interventions from a corpus of 64,177 research articles in the speech-language pathology domain, culminating in the creation of a publicly accessible intervention knowledge base with great potential to benefit the speech-language pathology community.

Yanbiao Ji, Qiuchang Li, Yuting Hu et al.

This paper introduces EnergyFlow, a framework that unifies generative action modeling with inverse reinforcement learning by parameterizing a scalar energy function whose gradient is the denoising field. We establish that under maximum-entropy optimality, the score function learned via denoising score matching recovers the gradient of the expert's soft Q-function, enabling reward extraction without adversarial training. Formally, we prove that constraining the learned field to be conservative reduces hypothesis complexity and tightens out-of-distribution generalization bounds. We further characterize the identifiability of recovered rewards and bound how score estimation errors propagate to action preferences. Empirically, EnergyFlow achieves state-of-the-art imitation performance on various manipulation tasks while providing an effective reward signal for downstream reinforcement learning that outperforms both adversarial IRL methods and likelihood-based alternatives. These results show that the structural constraints required for valid reward extraction simultaneously serve as beneficial inductive biases for policy generalization. The code is available at https://github.com/sotaagi/EnergyFlow.

Yuting Hu, Lei Zhuang, Hua Xiang et al.

As integrated circuit (IC) dimensions shrink below the lithographic wavelength, optical lithography faces growing challenges from diffraction and process variability. Model-based optical proximity correction (OPC) and inverse lithography technique (ILT) remain indispensable but computationally expensive, requiring repeated simulations that limit scalability. Although deep learning has been applied to mask optimization, existing datasets often rely on synthetic layouts, disregard standard-cell hierarchy, and neglect the surrounding contexts around the mask optimization targets, thereby constraining their applicability to practical mask optimization. To advance deep learning for cell- and context-aware mask optimization, we present MaskOpt, a large-scale benchmark dataset constructed from real IC designs at the 45$\mathrm{nm}$ node. MaskOpt includes 104,714 metal-layer tiles and 121,952 via-layer tiles. Each tile is clipped at a standard-cell placement to preserve cell information, exploiting repeated logic gate occurrences. Different context window sizes are supported in MaskOpt to capture the influence of neighboring shapes from optical proximity effects. We evaluate state-of-the-art deep learning models for IC mask optimization to build up benchmarks, and the evaluation results expose distinct trade-offs across baseline models. Further context size analysis and input ablation studies confirm the importance of both surrounding geometries and cell-aware inputs in achieving accurate mask generation.

Chenhui Xu, Dancheng Liu, Yuting Hu et al.

Physics-Informed Neural Networks (PINNs) are a kind of deep-learning-based numerical solvers for partial differential equations (PDEs). Existing PINNs often suffer from failure modes of being unable to propagate patterns of initial conditions. We discover that these failure modes are caused by the simplicity bias of neural networks and the mismatch between PDE's continuity and PINN's discrete sampling. We reveal that the State Space Model (SSM) can be a continuous-discrete articulation allowing initial condition propagation, and that simplicity bias can be eliminated by aligning a sequence of moderate granularity. Accordingly, we propose PINNMamba, a novel framework that introduces sub-sequence modeling with SSM. Experimental results show that PINNMamba can reduce errors by up to 86.3\% compared with state-of-the-art architecture. Our code is available at https://github.com/miniHuiHui/PINNMamba.

Ruiyang Qin, Yuting Hu, Zheyu Yan et al.

Neural Architecture Search (NAS) has become the de fecto tools in the industry in automating the design of deep neural networks for various applications, especially those driven by mobile and edge devices with limited computing resources. The emerging large language models (LLMs), due to their prowess, have also been incorporated into NAS recently and show some promising results. This paper conducts further exploration in this direction by considering three important design metrics simultaneously, i.e., model accuracy, fairness, and hardware deployment efficiency. We propose a novel LLM-based NAS framework, FL-NAS, in this paper, and show experimentally that FL-NAS can indeed find high-performing DNNs, beating state-of-the-art DNN models by orders-of-magnitude across almost all design considerations.

Yuting Hu, Lei Zhuang, Chen Wang et al.

As feature sizes shrink to the nanometer scale, accurately transferring circuit patterns from photomasks to silicon wafers becomes increasingly challenging. Optical proximity correction (OPC) is widely used to ensure pattern fidelity and manufacturability. Recent generative mask optimization models based on encoder-decoder architecture can synthesize near-optimal masks, serving as fast machine learning (ML) surrogates for traditional OPC. However, these models often fail to capture the geometric transformations from target layouts to mask patterns, leading to suboptimal quality. In this work, we formulate mask generation as a sequence of morphological operations on local layout features and propose \textit{MorphOPC}, a multi-scale hierarchical model with neural morphological modules to learn these transformations. Experiments on edge-based OPC and ILT benchmarks across metal and via layers show that \textit{MorphOPC} consistently outperforms state-of-the-art methods, achieving higher printing fidelity and lower manufacturing cost, demonstrating strong potential for scalable mask optimization.

Ruiyang Qin, Dancheng Liu, Gelei Xu et al.

The combination of Large Language Models (LLM) and Automatic Speech Recognition (ASR), when deployed on edge devices (called edge ASR-LLM), can serve as a powerful personalized assistant to enable audio-based interaction for users. Compared to text-based interaction, edge ASR-LLM allows accessible and natural audio interactions. Unfortunately, existing ASR-LLM models are mainly trained in high-performance computing environments and produce substantial model weights, making them difficult to deploy on edge devices. More importantly, to better serve users' personalized needs, the ASR-LLM must be able to learn from each distinct user, given that audio input often contains highly personalized characteristics that necessitate personalized on-device training. Since individually fine-tuning the ASR or LLM often leads to suboptimal results due to modality-specific limitations, end-to-end training ensures seamless integration of audio features and language understanding (cross-modal alignment), ultimately enabling a more personalized and efficient adaptation on edge devices. However, due to the complex training requirements and substantial computational demands of existing approaches, cross-modal alignment between ASR audio and LLM can be challenging on edge devices. In this work, we propose a resource-efficient cross-modal alignment framework that bridges ASR and LLMs on edge devices to handle personalized audio input. Our framework enables efficient ASR-LLM alignment on resource-constrained devices like NVIDIA Jetson Orin (8GB RAM), achieving 50x training time speedup while improving the alignment quality by more than 50\%. To the best of our knowledge, this is the first work to study efficient ASR-LLM alignment on resource-constrained edge devices.

Yuting Hu, Chenhui Xu, Ruiyang Qin et al.

Partial perception deficits can compromise autonomous vehicle safety by disrupting environmental understanding. Current protocols typically respond with immediate stops or minimal-risk maneuvers, worsening traffic flow and lacking flexibility for rare driving scenarios. In this paper, we propose LLM-RCO, a framework leveraging large language models to integrate human-like driving commonsense into autonomous systems facing perception deficits. LLM-RCO features four key modules: hazard inference, short-term motion planner, action condition verifier, and safety constraint generator. These modules interact with the dynamic driving environment, enabling proactive and context-aware control actions to override the original control policy of autonomous agents. To improve safety in such challenging conditions, we construct DriveLM-Deficit, a dataset of 53,895 video clips featuring deficits of safety-critical objects, complete with annotations for LLM-based hazard inference and motion planning fine-tuning. Extensive experiments in adverse driving conditions with the CARLA simulator demonstrate that systems equipped with LLM-RCO significantly improve driving performance, highlighting its potential for enhancing autonomous driving resilience against adverse perception deficits. Our results also show that LLMs fine-tuned with DriveLM-Deficit can enable more proactive movements instead of conservative stops in the context of perception deficits.

Zhantao Yang, Huangji Wang, Ruili Feng et al.

Generating captions for long and complex videos is both critical and challenging, with significant implications for the growing fields of text-to-video generation and multi-modal understanding. One key challenge in long video captioning is accurately recognizing the same individuals who appear in different frames, which we refer to as the ID-Matching problem. Few prior works have focused on this important issue. Those that have, usually suffer from limited generalization and depend on point-wise matching, which limits their overall effectiveness. In this paper, unlike previous approaches, we build upon LVLMs to leverage their powerful priors. We aim to unlock the inherent ID-Matching capabilities within LVLMs themselves to enhance the ID-Matching performance of captions. Specifically, we first introduce a new benchmark for assessing the ID-Matching capabilities of video captions. Using this benchmark, we investigate LVLMs containing GPT-4o, revealing key insights that the performance of ID-Matching can be improved through two methods: 1) enhancing the usage of image information and 2) increasing the quantity of information of individual descriptions. Based on these insights, we propose a novel video captioning method called Recognizing Identities for Captioning Effectively (RICE). Extensive experiments including assessments of caption quality and ID-Matching performance, demonstrate the superiority of our approach. Notably, when implemented on GPT-4o, our RICE improves the precision of ID-Matching from 50% to 90% and improves the recall of ID-Matching from 15% to 80% compared to baseline. RICE makes it possible to continuously track different individuals in the captions of long videos.

Yuting Hu, Jinjun Xiong

A full power flow (PF) model is a complete representation of the physical power network. Traditional model-based methods rely on the full PF model to implement power flow analysis. In practice, however, some PF model parameters can be inaccurate or even unavailable due to the uncertainties or dynamics in the power systems. Moreover, because the power network keeps evolving with possibly changing topology, the generalizability of a PF model to different network sizes and typologies should be considered. In this paper, we propose a PF rebuild model based on graph attention networks (GAT) by constructing a new graph based on the real and imaginary parts of voltage at each bus. By comparing with two state-of-the-art PF rebuild models for different standard IEEE power system cases and their modified topology variants, we demonstrate the feasibility of our method. Experimental results show that our proposed model achieves better accuracy for a changing network and can generalize to different networks with less accuracy discount.

Shangwen Zhu, Qianyu Peng, Yuting Hu et al.

Flow-based generative models have achieved remarkable progress, with classifier-free guidance (CFG) becoming the standard for high-fidelity generation. However, the conventional practice of applying a strong, fixed guidance scale throughout inference is poorly suited for the rapid, few-step sampling required by modern applications. In this work, we uncover the root cause of this conflict: a fundamental sampling instability where the earliest steps are acutely sensitive to guidance. We trace this to a significant spike in the ratio of conditional to unconditional predictions--a spike that we prove to be an inherent property of the training data distribution itself, making it a almost inevitable challenge. Applying a high, static guidance value during this volatile initial phase leads to an exponential amplification of error, degrading image quality. To resolve this, we propose a simple, theoretically grounded, adaptive guidance schedule that automatically dampens the guidance scale at early steps based on the evolving ratio. Our method is lightweight, incurs no inference overhead, and is compatible with standard frameworks. Experiments across state-of-the-art image (SD3.5, Qwen-Image) and video (WAN2.1) models show our approach enables up to 3x faster sampling while maintaining or improving quality, robustness, and semantic alignment. Our findings highlight that adapting guidance to the sampling process, rather than fixing it, is critical for unlocking the full potential of fast, flow-based models.

Dancheng Liu, Amir Nassereldine, Ziming Yang et al.

Large language models (LLMs) have attracted significant attention for their exceptional abilities in various natural language processing tasks, but they suffer from hallucinations that will cause performance degradation. One promising solution to improve the LLMs' performance is to ask LLMs to revise their answer after generation, a technique known as self-correction. Among the two types of self-correction, intrinsic self-correction is considered a promising direction because it does not utilize external knowledge. However, recent works doubt the validity of LLM's ability to conduct intrinsic self-correction. In this paper, we present a novel perspective on the intrinsic self-correction capabilities of LLMs through theoretical analyses and empirical experiments. In addition, we identify two critical factors for successful self-correction: zero temperature and fair prompts. Leveraging these factors, we demonstrate that intrinsic self-correction ability is exhibited across multiple existing LLMs. Our findings offer insights into the fundamental theories underlying the self-correction behavior of LLMs and remark on the importance of unbiased prompts and zero temperature settings in harnessing their full potential.

Yuting Hu, Zhiling Long, Anirudha Sundaresan et al.

Tactile sensing or fabric hand plays a critical role in an individual's decision to buy a certain fabric from the range of available fabrics for a desired application. Therefore, textile and clothing manufacturers have long been in search of an objective method for assessing fabric hand, which can then be used to engineer fabrics with a desired hand. Recognizing textures and materials in real-world images has played an important role in object recognition and scene understanding. In this paper, we explore how to computationally characterize apparent or latent properties (e.g., surface smoothness) of materials, i.e., computational material surface characterization, which moves a step further beyond material recognition. We formulate the problem as a very fine-grained texture classification problem, and study how deep learning-based texture representation techniques can help tackle the task. We introduce a new, large-scale challenging microscopic material surface dataset (CoMMonS), geared towards an automated fabric quality assessment mechanism in an intelligent manufacturing system. We then conduct a comprehensive evaluation of state-of-the-art deep learning-based methods for texture classification using CoMMonS. Additionally, we propose a multi-level texture encoding and representation network (MuLTER), which simultaneously leverages low- and high-level features to maintain both texture details and spatial information in the texture representation. Our results show that, in comparison with the state-of-the-art deep texture descriptors, MuLTER yields higher accuracy not only on our CoMMonS dataset for material characterization, but also on established datasets such as MINC-2500 and GTOS-mobile for material recognition.

Yuting Hu, Zhen Wang, Ghassan AlRegib

In this paper, we present an efficient and distinctive local descriptor, namely block intensity and gradient difference (BIGD). In an image patch, we randomly sample multi-scale block pairs and utilize the intensity and gradient differences of pairwise blocks to construct the local BIGD descriptor. The random sampling strategy and the multi-scale framework help BIGD descriptors capture the distinctive patterns of patches at different orientations and spatial granularity levels. We use vectors of locally aggregated descriptors (VLAD) or improved Fisher vector (IFV) to encode local BIGD descriptors into a full image descriptor, which is then fed into a linear support vector machine (SVM) classifier for texture classification. We compare the proposed descriptor with typical and state-of-the-art ones by evaluating their classification performance on five public texture data sets including Brodatz, CUReT, KTH-TIPS, and KTH-TIPS-2a and -2b. Experimental results show that the proposed BIGD descriptor with stronger discriminative power yields 0.12% ~ 6.43% higher classification accuracy than the state-of-the-art texture descriptor, dense microblock difference (DMD).

Zhongliang Yang, Yuting Hu, Yongfeng Huang et al.

The purpose of the covert communication system is to implement the communication process without causing third party perception. In order to achieve complete covert communication, two aspects of security issues need to be considered. The first one is to cover up the existence of information, that is, to ensure the content security of information; the second one is to cover up the behavior of transmitting information, that is, to ensure the behavioral security of communication. However, most of the existing information hiding models are based on the "Prisoners' Model", which only considers the content security of carriers, while ignoring the behavioral security of the sender and receiver. We think that this is incomplete for the security of covert communication. In this paper, we propose a new covert communication framework, which considers both content security and behavioral security in the process of information transmission. In the experimental part, we analyzed a large amount of collected real Twitter data to illustrate the security risks that may be brought to covert communication if we only consider content security and neglect behavioral security. Finally, we designed a toy experiment, pointing out that in addition to most of the existing content steganography, under the proposed new framework of covert communication, we can also use user's behavior to implement behavioral steganography. We hope this new proposed framework will help researchers to design better covert communication systems.

Hasan Al-Marzouqi, Yuting Hu, Ghassan AlRegib

Image retrieval is an important problem in the area of multimedia processing. This paper presents two new curvelet-based algorithms for texture retrieval which are suitable for use in constrained-memory devices. The developed algorithms are tested on three publicly available texture datasets: CUReT, Mondial-Marmi, and STex-fabric. Our experiments confirm the effectiveness of the proposed system. Furthermore, a weighted version of the proposed retrieval algorithm is proposed, which is shown to achieve promising results in the classification of seismic activities.

Yuting Hu, Zhiling Long, Ghassan AlRegib

In this paper, we propose a multi-level texture encoding and representation network (MuLTER) for texture-related applications. Based on a multi-level pooling architecture, the MuLTER network simultaneously leverages low- and high-level features to maintain both texture details and spatial information. Such a pooling architecture involves few extra parameters and keeps feature dimensions fixed despite of the changes of image sizes. In comparison with state-of-the-art texture descriptors, the MuLTER network yields higher recognition accuracy on typical texture datasets such as MINC-2500 and GTOS-mobile with a discriminative and compact representation. In addition, we analyze the impact of combining features from different levels, which supports our claim that the fusion of multi-level features efficiently enhances recognition performance. Our source code will be published on GitHub (https://github.com/olivesgatech).

Zhongliang Yang, Hao Yang, Yuting Hu et al.

Previous VoIP steganalysis methods face great challenges in detecting speech signals at low embedding rates, and they are also generally difficult to perform real-time detection, making them hard to truly maintain cyberspace security. To solve these two challenges, in this paper, combined with the sliding window detection algorithm and Convolution Neural Network we propose a real-time VoIP steganalysis method which based on multi-channel convolution sliding windows. In order to analyze the correlations between frames and different neighborhood frames in a VoIP signal, we define multi channel sliding detection windows. Within each sliding window, we design two feature extraction channels which contain multiple convolution layers with multiple convolution kernels each layer to extract correlation features of the input signal. Then based on these extracted features, we use a forward fully connected network for feature fusion. Finally, by analyzing the statistical distribution of these features, the discriminator will determine whether the input speech signal contains covert information or not.We designed several experiments to test the proposed model's detection ability under various conditions, including different embedding rates, different speech length, etc. Experimental results showed that the proposed model outperforms all the previous methods, especially in the case of low embedding rate, which showed state-of-the-art performance. In addition, we also tested the detection efficiency of the proposed model, and the results showed that it can achieve almost real-time detection of VoIP speech signals.

Zhiling Long, Yazeed Alaudah, Muhammad Ali Qureshi et al.

In this paper, we explore how to computationally characterize subsurface geological structures presented in seismic volumes using texture attributes. For this purpose, we conduct a comparative study of typical texture attributes presented in the image processing literature. We focus on spatial attributes in this study and examine them in a new application for seismic interpretation, i.e., seismic volume labeling. For this application, a data volume is automatically segmented into various structures, each assigned with its corresponding label. If the labels are assigned with reasonable accuracy, such volume labeling will help initiate an interpretation process in a more effective manner. Our investigation proves the feasibility of accomplishing this task using texture attributes. Through the study, we also identify advantages and disadvantages associated with each attribute.

Yuting Hu, Liang Zheng, Yi Yang et al.

This paper contributes a new large-scale dataset for weakly supervised cross-media retrieval, named Twitter100k. Current datasets, such as Wikipedia, NUS Wide and Flickr30k, have two major limitations. First, these datasets are lacking in content diversity, i.e., only some pre-defined classes are covered. Second, texts in these datasets are written in well-organized language, leading to inconsistency with realistic applications. To overcome these drawbacks, the proposed Twitter100k dataset is characterized by two aspects: 1) it has 100,000 image-text pairs randomly crawled from Twitter and thus has no constraint in the image categories; 2) text in Twitter100k is written in informal language by the users. Since strongly supervised methods leverage the class labels that may be missing in practice, this paper focuses on weakly supervised learning for cross-media retrieval, in which only text-image pairs are exploited during training. We extensively benchmark the performance of four subspace learning methods and three variants of the Correspondence AutoEncoder, along with various text features on Wikipedia, Flickr30k and Twitter100k. Novel insights are provided. As a minor contribution, inspired by the characteristic of Twitter100k, we propose an OCR-based cross-media retrieval method. In experiment, we show that the proposed OCR-based method improves the baseline performance.