arXiv subjects

Jiaxing Shen

Publications and source records attributed to Jiaxing Shen.

At least 19 recordsLinked to original sources

arXiv2026

Reflection Steering: Disentangling Reflection from Reasoning in Activation Space for Token-Efficient Inference

Large reasoning models often produce reasoning traces with verification, revision, and backtracking. When reflection merely re-checks established results, it wastes reasoning tokens and increases latency. Most existing reflection steering methods add a label-derived mean-difference direction across preset layers, but its entanglement with reasoning and length signals destabilizes the accuracy-efficiency trade-off. In this paper, we propose Reflection Steering, a training-free framework for controlling reflection-associated computation within LLMs by disentangling reflection-related activations from general reasoning. Specifically, we contrast reflective and non-reflective hidden states at each LLM layer, denoise the resulting reflection directions with PCA, and orthogonalize them against general-reasoning directions. To limit downstream amplification from early-layer interventions, we calibrate each layer across multiple intervention strengths on a small set, retain only stable layers, and apply bounded projection removal to their residual-stream activations. We conduct extensive experiments across two public benchmarks and three open-weight LLMs against state-of-the-art activation-steering baselines. Results show that Reflection Steering reduces reasoning tokens by 16.9% on average across six matched settings. Besides, our method further introduces a bounded reflection intervention-strength parameter $α$, enabling deployment-time adjustment to balance token savings, accuracy, and generation stability.

cs.LG↗

arXiv2026

Reprojection-Guided 3D Gaussian Splatting Diffusion for Weakly Supervised Single-Image Normal Estimation

We propose CLONE, a Continuous Latent Optimization framework for Normal Estimation via 3D Gaussian splatting. The core idea is to construct an image-geometry-image consistency strategy that unifies explicit geometric representation with differentiable rendering, thereby enabling weakly supervised learning without normal ground truth. Specifically, CLONE comprises four components. First, by introducing a differentiable light interaction model with a learnable modulation kernel, we perform a unified reparameterization of the 3DGS parameter space. Second, the conditional single-step deterministic refinement network integrates denoising architectures with differentiable reprojection constraints to refine the initial normals, thereby adaptively recovering the high-frequency details erased by the inherently smooth Gaussian primitives. Third, the cross-domain gating fusion mechanism adaptively combines the two complementary normal estimates, reconciling the geometrically consistent yet over-smooth 3DGS estimate with the detailed yet potentially geometry-inconsistent refinement. Finally, all components are jointly optimized under a unified photometric reprojection objective with geometric consistency regularizations in a fully differentiable pathway, achieving an end-to-end optimization closed loop without relying on external normal labels.

cs.CV↗

arXiv2025

Mind the Gap: Revealing Inconsistencies Across Heterogeneous AI Accelerators

While NVIDIA remains the dominant provider of AI accelerators within cloud data center, emerging vendors such as AMD, Intel, Mac, and Huawei offer cost-effective alternatives with claims of compatibility and performance. This paper presents the first empirical study investigating divergence in machine learning model across heterogeneous AI accelerators. Utilizing an automated pipeline, we synthesize over 100,000 variant models derived from 4,000 real-world models and execute them across five different enterprise-grade accelerators. Our findings suggest that newer AI platforms from Mac and Huawei support at least 17\% fewer operators than NVIDIA. These platforms also exhibit a higher rate of output discrepancies (exceeding 5\%), which stem from differences in operator implementations, handling of exceptional numerical values, and instruction scheduling. They are also more susceptible to failures during model compilation-based acceleration, and in some cases, the compiled models produce outputs that differ noticeably from those generated using the standard execution mode. In addition, we identify 7 implementation flaws in PyTorch and 40 platform-specific issues across vendors. These results underscore the challenges of achieving consistent machine learning behavior in an increasingly diverse hardware ecosystem.

cs.DC↗

arXiv2025

SGSimEval: A Comprehensive Multifaceted and Similarity-Enhanced Benchmark for Automatic Survey Generation Systems

The growing interest in automatic survey generation (ASG), a task that traditionally required considerable time and effort, has been spurred by recent advances in large language models (LLMs). With advancements in retrieval-augmented generation (RAG) and the rising popularity of multi-agent systems (MASs), synthesizing academic surveys using LLMs has become a viable approach, thereby elevating the need for robust evaluation methods in this domain. However, existing evaluation methods suffer from several limitations, including biased metrics, a lack of human preference, and an over-reliance on LLMs-as-judges. To address these challenges, we propose SGSimEval, a comprehensive benchmark for Survey Generation with Similarity-Enhanced Evaluation that evaluates automatic survey generation systems by integrating assessments of the outline, content, and references, and also combines LLM-based scoring with quantitative metrics to provide a multifaceted evaluation framework. In SGSimEval, we also introduce human preference metrics that emphasize both inherent quality and similarity to humans. Extensive experiments reveal that current ASG systems demonstrate human-comparable superiority in outline generation, while showing significant room for improvement in content and reference generation, and our evaluation metrics maintain strong consistency with human assessments.

cs.CL↗

arXiv2025

Inverse Design of Metamaterials with Manufacturing-Guiding Spectrum-to-Structure Conditional Diffusion Model

Metamaterials are artificially engineered structures that manipulate electromagnetic waves, having optical properties absent in natural materials. Recently, machine learning for the inverse design of metamaterials has drawn attention. However, the highly nonlinear relationship between the metamaterial structures and optical behaviour, coupled with fabrication difficulties, poses challenges for using machine learning to design and manufacture complex metamaterials. Herein, we propose a general framework that implements customised spectrum-to-shape and size parameters to address one-to-many metamaterial inverse design problems using conditional diffusion models. Our method exhibits superior spectral prediction accuracy, generates a diverse range of patterns compared to other typical generative models, and offers valuable prior knowledge for manufacturing through the subsequent analysis of the diverse generated results, thereby facilitating the experimental fabrication of metamaterial designs. We demonstrate the efficacy of the proposed method by successfully designing and fabricating a free-form metamaterial with a tailored selective emission spectrum for thermal camouflage applications.

physics.optics↗

arXiv2024

FairTP: A Prolonged Fairness Framework for Traffic Prediction

Traffic prediction plays a crucial role in intelligent transportation systems. Existing approaches primarily focus on improving overall accuracy, often neglecting a critical issue: whether predictive models lead to biased decisions by transportation authorities. In practice, the uneven deployment of traffic sensors across urban areas results in imbalanced data, causing prediction models to perform poorly in certain regions and leading to unfair decision-making. This imbalance ultimately harms the equity and quality of life for residents. Moreover, current fairness-aware machine learning models only ensure fairness at specific time points, failing to maintain fairness over extended periods. As traffic conditions change, such static fairness approaches become ineffective. To address this gap, we propose FairTP, a framework for prolonged fair traffic prediction. We introduce two new fairness definitions tailored for dynamic traffic scenarios. Fairness in traffic prediction is not static; it varies over time and across regions. Each sensor or urban area can alternate between two states: "sacrifice" (low prediction accuracy) and "benefit" (high prediction accuracy). Prolonged fairness is achieved when the overall states of sensors remain similar over a given period. We define two types of fairness: region-based static fairness and sensor-based dynamic fairness. To implement this, FairTP incorporates a state identification module to classify sensors' states as either "sacrifice" or "benefit," enabling prolonged fairness-aware predictions. Additionally, we introduce a state-guided balanced sampling strategy to further enhance fairness, addressing performance disparities among regions with uneven sensor distributions. Extensive experiments on two real-world datasets demonstrate that FairTP significantly improves prediction fairness while minimizing accuracy degradation.

cs.LG↗

arXiv2024

Prompting Explicit and Implicit Knowledge for Multi-hop Question Answering Based on Human Reading Process

Pre-trained language models (PLMs) leverage chains-of-thought (CoT) to simulate human reasoning and inference processes, achieving proficient performance in multi-hop QA. However, a gap persists between PLMs' reasoning abilities and those of humans when tackling complex problems. Psychological studies suggest a vital connection between explicit information in passages and human prior knowledge during reading. Nevertheless, current research has given insufficient attention to linking input passages and PLMs' pre-training-based knowledge from the perspective of human cognition studies. In this study, we introduce a Prompting Explicit and Implicit knowledge (PEI) framework, which uses prompts to connect explicit and implicit knowledge, aligning with human reading process for multi-hop QA. We consider the input passages as explicit knowledge, employing them to elicit implicit knowledge through unified prompt reasoning. Furthermore, our model incorporates type-specific reasoning via prompts, a form of implicit knowledge. Experimental results show that PEI performs comparably to the state-of-the-art on HotpotQA. Ablation studies confirm the efficacy of our model in bridging and integrating explicit and implicit knowledge.

cs.CL↗

arXiv2024

Keep Me Updated: An Empirical Study of Proprietary Vendor Blobs in Android Firmware

Despite extensive security research on various Android components, such as kernel or runtime, little attention has been paid to the proprietary vendor blobs within Android firmware. In this paper, we conduct a large-scale empirical study to understand the update patterns and assess the security implications of vendor blobs. We specifically focus on GPU blobs because they are loaded into every process for displaying graphics user interfaces and can affect the entire system's security. We examine over 13,000 Android firmware releases between January 2018 and April 2024. Our results reveal that device manufacturers often neglect vendor blob updates. About 82\% of firmware releases contain outdated GPU blobs (up to 1,281 days). A significant number of blobs also rely on obsolete LLVM core libraries released more than 15 years ago. To analyze their security implications, we develop a performant fuzzer that requires no physical access to mobile devices. We discover 289 security and behavioral bugs within the blobs. We also present a case study demonstrating how these vulnerabilities can be exploited via WebGL. This work underscores the critical security concerns associated with vulnerable vendor blobs and emphasizes the urgent need for timely updates from device manufacturers.

cs.SE↗

arXiv2024

ViDSOD-100: A New Dataset and a Baseline Model for RGB-D Video Salient Object Detection

With the rapid development of depth sensor, more and more RGB-D videos could be obtained. Identifying the foreground in RGB-D videos is a fundamental and important task. However, the existing salient object detection (SOD) works only focus on either static RGB-D images or RGB videos, ignoring the collaborating of RGB-D and video information. In this paper, we first collect a new annotated RGB-D video SOD (ViDSOD-100) dataset, which contains 100 videos within a total of 9,362 frames, acquired from diverse natural scenes. All the frames in each video are manually annotated to a high-quality saliency annotation. Moreover, we propose a new baseline model, named attentive triple-fusion network (ATF-Net), for RGB-D video salient object detection. Our method aggregates the appearance information from an input RGB image, spatio-temporal information from an estimated motion map, and the geometry information from the depth map by devising three modality-specific branches and a multi-modality integration branch. The modality-specific branches extract the representation of different inputs, while the multi-modality integration branch combines the multi-level modality-specific features by introducing the encoder feature aggregation (MEA) modules and decoder feature aggregation (MDA) modules. The experimental findings conducted on both our newly introduced ViDSOD-100 dataset and the well-established DAVSOD dataset highlight the superior performance of the proposed ATF-Net. This performance enhancement is demonstrated both quantitatively and qualitatively, surpassing the capabilities of current state-of-the-art techniques across various domains, including RGB-D saliency detection, video saliency detection, and video object segmentation. Our data and our code are available at github.com/jhl-Det/RGBD_Video_SOD.

cs.CV↗

arXiv2024

Dynamic hysteresis of an oscillatory contact line

During oscillatory wetting, a phase retardation emerges between contact angle variation and contact line velocity, presenting as a hysteresis loop in their correlation -- an effect we term dynamic hysteresis. This phenomenon is found to be tunable by modifying the surface with different molecular layers. A comparative analysis of dynamic hysteresis, static hysteresis, and contact line friction coefficients across diverse substrates reveals that dynamic hysteresis is not a result of dissipative effects but is instead proportionally linked to the flexibility of the grafted layer on the surface. In the quest for appropriate conditions to model oscillatory contact line motion, we identify the generalized Hocking's linear law and modified Generalized Navier Boundary Condition (GNBC) as alternative options for predicting realistic dynamic hysteresis.

physics.flu-dyn↗

arXiv2024

Private Data Leakage in Federated Human Activity Recognition for Wearable Healthcare Devices

Wearable data serves various health monitoring purposes, such as determining activity states based on user behavior and providing tailored exercise recommendations. However, the individual data perception and computational capabilities of wearable devices are limited, often necessitating the joint training of models across multiple devices. Federated Human Activity Recognition (HAR) presents a viable research avenue, allowing for global model training without the need to upload users' local activity data. Nonetheless, recent studies have revealed significant privacy concerns persisting within federated learning frameworks. To address this gap, we focus on investigating privacy leakage issues within federated user behavior recognition modeling across multiple wearable devices. Our proposed system entails a federated learning architecture comprising $N$ wearable device users and a parameter server, which may exhibit curiosity in extracting sensitive user information from model parameters. Consequently, we consider a membership inference attack based on a malicious server, leveraging differences in model generalization across client data. Experimentation conducted on five publicly available HAR datasets demonstrates an accuracy rate of 92\% for malicious server-based membership inference. Our study provides preliminary evidence of substantial privacy risks associated with federated training across multiple wearable devices, offering a novel research perspective within this domain.

cs.CR↗

arXiv2024

Personality-affected Emotion Generation in Dialog Systems

Generating appropriate emotions for responses is essential for dialog systems to provide human-like interaction in various application scenarios. Most previous dialog systems tried to achieve this goal by learning empathetic manners from anonymous conversational data. However, emotional responses generated by those methods may be inconsistent, which will decrease user engagement and service quality. Psychological findings suggest that the emotional expressions of humans are rooted in personality traits. Therefore, we propose a new task, Personality-affected Emotion Generation, to generate emotion based on the personality given to the dialog system and further investigate a solution through the personality-affected mood transition. Specifically, we first construct a daily dialog dataset, Personality EmotionLines Dataset (PELD), with emotion and personality annotations. Subsequently, we analyze the challenges in this task, i.e., (1) heterogeneously integrating personality and emotional factors and (2) extracting multi-granularity emotional information in the dialog context. Finally, we propose to model the personality as the transition weight by simulating the mood transition process in the dialog system and solve the challenges above. We conduct extensive experiments on PELD for evaluation. Results suggest that by adopting our method, the emotion generation performance is improved by 13% in macro-F1 and 5% in weighted-F1 from the BERT-base model.

cs.CL↗

arXiv2024

PreConfig: A Pretrained Model for Automating Network Configuration

Manual network configuration automation (NCA) tools face significant challenges in versatility and flexibility due to their reliance on extensive domain expertise and manual design, limiting their adaptability to diverse scenarios and complex application needs. This paper introduces PreConfig, an innovative NCA tool that leverages a pretrained language model for automating network configuration tasks. PreConfig is designed to address the complexity and variety of NCA tasks by framing them as text-to-text transformation problems, thus unifying the tasks of configuration generation, translation, and analysis under a single, versatile model. Our approach overcomes existing tools' limitations by utilizing advances in natural language processing to automatically comprehend and generate network configurations without extensive manual re-engineering. We confront the challenges of integrating domain-specific knowledge into pretrained models and the scarcity of supervision data in the network configuration field. Our solution involves constructing a specialized corpus and further pretraining on network configuration data, coupled with a novel data mining technique for generating task supervision data. The proposed model demonstrates robustness in configuration generation, translation, and analysis, outperforming conventional tools in handling complex networking environments. The experimental results validate the effectiveness of PreConfig, establishing a new direction for automating network configuration tasks with pretrained language models.

cs.NI↗

arXiv2024

Machine Unlearning in Large Language Models

Recently, large language models (LLMs) have emerged as a notable field, attracting significant attention for its ability to automatically generate intelligent contents for various application domains. However, LLMs still suffer from significant security and privacy issues. For example, LLMs might expose user privacy from hacking attacks or targeted prompts. To address this problem, this paper introduces a novel machine unlearning framework into LLMs. Our objectives are to make LLMs not produce harmful, hallucinatory, or privacy-compromising responses, while retaining their standard output capabilities. To accomplish this, we use an evaluative model to pinpoint dialogues needing unlearning. We also establish a distance loss to function as the model's negative loss, diverting it from previous undesirable outputs. Furthermore, we determine the expected output's cluster mean to formulate a positive loss, directing the model's outputs toward preferable outcomes without compromising its reasoning abilities and performance. Experimental results show that our approach effectively meets unlearning objectives without substantially compromising model performance.

cs.CR↗

arXiv2024

Federated Unlearning for Human Activity Recognition

The rapid evolution of Internet of Things (IoT) technology has spurred the widespread adoption of Human Activity Recognition (HAR) in various daily life domains. Federated Learning (FL) is frequently utilized to build a global HAR model by aggregating user contributions without transmitting raw individual data. Despite substantial progress in user privacy protection with FL, challenges persist. Regulations like the General Data Protection Regulation (GDPR) empower users to request data removal, raising a new query in FL: How can a HAR client request data removal without compromising other clients' privacy? In response, we propose a lightweight machine unlearning method for refining the FL HAR model by selectively removing a portion of a client's training data. Our method employs a third-party dataset unrelated to model training. Using KL divergence as a loss function for fine-tuning, we aim to align the predicted probability distribution on forgotten data with the third-party dataset. Additionally, we introduce a membership inference evaluation method to assess unlearning effectiveness. Experimental results across diverse datasets show our method achieves unlearning accuracy comparable to \textit{retraining} methods, resulting in speedups ranging from hundreds to thousands.

cs.LG↗

arXiv2023

BAGEL: Backdoor Attacks against Federated Contrastive Learning

Federated Contrastive Learning (FCL) is an emerging privacy-preserving paradigm in distributed learning for unlabeled data. In FCL, distributed parties collaboratively learn a global encoder with unlabeled data, and the global encoder could be widely used as a feature extractor to build models for many downstream tasks. However, FCL is also vulnerable to many security threats (e.g., backdoor attacks) due to its distributed nature, which are seldom investigated in existing solutions. In this paper, we study the backdoor attack against FCL as a pioneer research, to illustrate how backdoor attacks on distributed local clients act on downstream tasks. Specifically, in our system, malicious clients can successfully inject a backdoor into the global encoder by uploading poisoned local updates, thus downstream models built with this global encoder will also inherit the backdoor. We also investigate how to inject backdoors into multiple downstream models, in terms of two different backdoor attacks, namely the \textit{centralized attack} and the \textit{decentralized attack}. Experiment results show that both the centralized and the decentralized attacks can inject backdoors into downstream models effectively with high attack success rates. Finally, we evaluate two defense methods against our proposed backdoor attacks in FCL, which indicates that the decentralized backdoor attack is more stealthy and harder to defend.

cs.CR↗

arXiv2023

Take Your Pick: Enabling Effective Personalized Federated Learning within Low-dimensional Feature Space

Personalized federated learning (PFL) is a popular framework that allows clients to have different models to address application scenarios where clients' data are in different domains. The typical model of a client in PFL features a global encoder trained by all clients to extract universal features from the raw data and personalized layers (e.g., a classifier) trained using the client's local data. Nonetheless, due to the differences between the data distributions of different clients (aka, domain gaps), the universal features produced by the global encoder largely encompass numerous components irrelevant to a certain client's local task. Some recent PFL methods address the above problem by personalizing specific parameters within the encoder. However, these methods encounter substantial challenges attributed to the high dimensionality and non-linearity of neural network parameter space. In contrast, the feature space exhibits a lower dimensionality, providing greater intuitiveness and interpretability as compared to the parameter space. To this end, we propose a novel PFL framework named FedPick. FedPick achieves PFL in the low-dimensional feature space by selecting task-relevant features adaptively for each client from the features generated by the global encoder based on its local data distribution. It presents a more accessible and interpretable implementation of PFL compared to those methods working in the parameter space. Extensive experimental results show that FedPick could effectively select task-relevant features for each client and improve model performance in cross-domain FL.

cs.LG↗

arXiv2023

Unlocking the Potential of Federated Learning for Deeper Models

Federated learning (FL) is a new paradigm for distributed machine learning that allows a global model to be trained across multiple clients without compromising their privacy. Although FL has demonstrated remarkable success in various scenarios, recent studies mainly utilize shallow and small neural networks. In our research, we discover a significant performance decline when applying the existing FL framework to deeper neural networks, even when client data are independently and identically distributed (i.i.d.). Our further investigation shows that the decline is due to the continuous accumulation of dissimilarities among client models during the layer-by-layer back-propagation process, which we refer to as "divergence accumulation." As deeper models involve a longer chain of divergence accumulation, they tend to manifest greater divergence, subsequently leading to performance decline. Both theoretical derivations and empirical evidence are proposed to support the existence of divergence accumulation and its amplified effects in deeper models. To address this issue, we propose several technical guidelines based on reducing divergence, such as using wider models and reducing the receptive field. These approaches can greatly improve the accuracy of FL on deeper models. For example, the application of these guidelines can boost the ResNet101 model's performance by as much as 43\% on the Tiny-ImageNet dataset.

cs.LG↗