Search

Recently, emotional speech generation and speaker cloning have garnered significant interest in text-to-speech (TTS). With the open-sourcing of codec language TTS models trained on massive datasets with large-scale parameters, adapting these general pre-trained TTS models to generate speech with specific emotional expressions and target speaker characteristics has become a topic of great attention. Common approaches, such as full and adapter-based fine-tuning, often overlook the specific contributions of model parameters to emotion and speaker control. Treating all parameters uniformly during fine-tuning, especially when the target data has limited content diversity compared to the pre-training corpus, results in slow training speed and an increased risk of catastrophic forgetting. To address these challenges, we propose a characteristic-specific partial fine-tuning strategy, short as CSP-FT. First, we use a weighted-sum approach to analyze the contributions of different Transformer layers in a pre-trained codec language TTS model for emotion and speaker control in the generated speech. We then selectively fine-tune the layers with the highest and lowest characteristic-specific contributions to generate speech with target emotional expression and speaker identity. Experimental results demonstrate that our method achieves performance comparable to, or even surpassing, full fine-tuning in generating speech with specific emotional expressions and speaker identities. Additionally, CSP-FT delivers approximately 2x faster training speeds, fine-tunes only around 8% of parameters, and significantly reduces catastrophic forgetting. Furthermore, we show that codec language TTS models perform competitively with self-supervised models in speaker identification and emotion classification tasks, offering valuable insights for developing universal speech processing models., Comment: 13 pages

Version control systems are commonly used to manage open-source software, in which each commit may introduce new vulnerabilities or fix existing ones. Researchers have developed various tools for detecting vulnerabilities in code commits, but their performance is limited by factors such as neglecting descriptive data and challenges in accurately identifying vulnerability introductions. To overcome these limitations, we propose CommitShield, which combines the code analysis capabilities of static analysis tools with the natural language and code understanding capabilities of large language models (LLMs) to enhance the accuracy of vulnerability introduction and fix detection by generating precise descriptions and obtaining rich patch contexts. We evaluate CommitShield using the newly constructed vulnerability repair dataset, CommitVulFix, and a cleaned vulnerability introduction dataset. Experimental results indicate that CommitShield improves recall by 76%-87% over state-of-the-art methods in the vulnerability fix detection task, and its F1-score improves by 15%-27% in the vulnerability introduction detection task.

The enhanced Gaussian noise (EGN) model, which accounts for inter-channel stimulated Raman scattering (ISRS), has been extensively utilized for evaluating nonlinear interference (NLI) within the C+L band. Compared to closed-form expressions and machine learning-based NLI evaluation models, it demonstrates broader applicability and its accuracy is not dependent on the support of large-scale datasets. However, its high computational complexity often results in lengthy computation times. Through analysis, the high-frequency oscillations of the four-wave mixing (FWM) efficiency factor integrand were identified as a primary factor limiting the computational speed of the ISRS EGN model. To address this issue, we propose an accurate approximation method that enables the derivation of a closed-form expression for the FWM efficiency factor without imposing restrictive conditions. Thereby, the scheme proposed in this paper could significantly accelerate the computational speed. Numerical results demonstrate that method in this work could achieve low error levels under high ISRS influence levels, with an MAE of less than 0.001 dB, and no cumulative error over increasing transmission distances, while reducing computation time by over 97%. Furthermore, a parallel computation strategy targeting independent regions within the integration domain is proposed, which could further improve computational efficiency by nearly 11 times., Comment: 12 pages, 12 figures, preprint submitted to IEEE for possible publication

Pre-training backbone networks on a general annotated dataset (e.g., ImageNet) that comprises numerous manually collected images with category annotations has proven to be indispensable for enhancing the generalization capacity of downstream visual tasks. However, those manually collected images often exhibit bias, which is non-transferable across either categories or domains, thus causing the model's generalization capacity degeneration. To mitigate this problem, we present an unbiased general annotated dataset generation framework (ubGen). Instead of expensive manual collection, we aim at directly generating unbiased images with category annotations. To achieve this goal, we propose to leverage the advantage of a multimodal foundation model (e.g., CLIP), in terms of aligning images in an unbiased semantic space defined by language. Specifically, we develop a bi-level semantic alignment loss, which not only forces all generated images to be consistent with the semantic distribution of all categories belonging to the target dataset in an adversarial learning manner, but also requires each generated image to match the semantic description of its category name. In addition, we further cast an existing image quality scoring model into a quality assurance loss to preserve the quality of the generated image. By leveraging these two loss functions, we can obtain an unbiased image generation model by simply fine-tuning a pre-trained diffusion model using only all category names in the target dataset as input. Experimental results confirm that, compared with the manually labeled dataset or other synthetic datasets, the utilization of our generated unbiased datasets leads to stable generalization capacity enhancement of different backbone networks across various tasks, especially in tasks where the manually labeled samples are scarce., Comment: Preprint

Generating sound effects for product-level videos, where only a small amount of labeled data is available for diverse scenes, requires the production of high-quality sounds in few-shot settings. To tackle the challenge of limited labeled data in real-world scenes, we introduce YingSound, a foundation model designed for video-guided sound generation that supports high-quality audio generation in few-shot settings. Specifically, YingSound consists of two major modules. The first module uses a conditional flow matching transformer to achieve effective semantic alignment in sound generation across audio and visual modalities. This module aims to build a learnable audio-visual aggregator (AVA) that integrates high-resolution visual features with corresponding audio features at multiple stages. The second module is developed with a proposed multi-modal visual-audio chain-of-thought (CoT) approach to generate finer sound effects in few-shot settings. Finally, an industry-standard video-to-audio (V2A) dataset that encompasses various real-world scenarios is presented. We show that YingSound effectively generates high-quality synchronized sounds across diverse conditional inputs through automated evaluations and human studies. Project Page: \url{https://giantailab.github.io/yingsound/}, Comment: 16 pages, 4 figures

Large Language Models (LLMs) have transformed numerous fields by enabling advanced natural language interactions but remain susceptible to critical vulnerabilities, particularly jailbreak attacks. Current jailbreak techniques, while effective, often depend on input modifications, making them detectable and limiting their stealth and scalability. This paper presents Targeted Model Editing (TME), a novel white-box approach that bypasses safety filters by minimally altering internal model structures while preserving the model's intended functionalities. TME identifies and removes safety-critical transformations (SCTs) embedded in model matrices, enabling malicious queries to bypass restrictions without input modifications. By analyzing distinct activation patterns between safe and unsafe queries, TME isolates and approximates SCTs through an optimization process. Implemented in the D-LLM framework, our method achieves an average Attack Success Rate (ASR) of 84.86% on four mainstream open-source LLMs, maintaining high performance. Unlike existing methods, D-LLM eliminates the need for specific triggers or harmful response collections, offering a stealthier and more effective jailbreak strategy. This work reveals a covert and robust threat vector in LLM security and emphasizes the need for stronger safeguards in model safety alignment.

Contemporary Text-to-Image (T2I) models frequently depend on qualitative human evaluations to assess the consistency between synthesized images and the text prompts. There is a demand for quantitative and automatic evaluation tools, given that human evaluation lacks reproducibility. We believe that an effective T2I evaluation metric should accomplish the following: detect instances where the generated images do not align with the textual prompts, a discrepancy we define as the `hallucination problem' in T2I tasks; record the types and frequency of hallucination issues, aiding users in understanding the causes of errors; and provide a comprehensive and intuitive scoring that close to human standard. To achieve these objectives, we propose a method based on large language models (LLMs) for conducting question-answering with an extracted scene-graph and created a dataset with human-rated scores for generated images. From the methodology perspective, we combine knowledge-enhanced question-answering tasks with image evaluation tasks, making the evaluation metrics more controllable and easier to interpret. For the contribution on the dataset side, we generated 12,000 synthesized images based on 1,000 composited prompts using three advanced T2I models. Subsequently, we conduct human scoring on all synthesized images and prompt pairs to validate the accuracy and effectiveness of our method as an evaluation metric. All generated images and the human-labeled scores will be made publicly available in the future to facilitate ongoing research on this crucial issue. Extensive experiments show that our method aligns more closely with human scoring patterns than other evaluation metrics.

To detect prohibited items in challenging categories, human inspectors typically rely on images from two distinct views (vertical and side). Can AI detect prohibited items from dual-view X-ray images in the same way humans do? Existing X-ray datasets often suffer from limitations, such as single-view imaging or insufficient sample diversity. To address these gaps, we introduce the Large-scale Dual-view X-ray (LDXray), which consists of 353,646 instances across 12 categories, providing a diverse and comprehensive resource for training and evaluating models. To emulate human intelligence in dual-view detection, we propose the Auxiliary-view Enhanced Network (AENet), a novel detection framework that leverages both the main and auxiliary views of the same object. The main-view pipeline focuses on detecting common categories, while the auxiliary-view pipeline handles more challenging categories using ``expert models" learned from the main view. Extensive experiments on the LDXray dataset demonstrate that the dual-view mechanism significantly enhances detection performance, e.g., achieving improvements of up to 24.7% for the challenging category of umbrellas. Furthermore, our results show that AENet exhibits strong generalization across seven different detection models for X-ray Inspection, Comment: 10 pages, 6 figures

The detection of prohibited items in X-ray security inspections is vital for ensuring public safety. However, the long-tail distribution of item categories, where certain prohibited items are far less common, poses a big challenge for detection models, as rare categories often lack sufficient training data. Existing methods struggle to classify these rare items accurately due to this imbalance. In this paper, we propose a Dual-level Boost Network (DBNet) specifically designed to overcome these challenges in X-ray security screening. Our approach introduces two key innovations: (1) a specific data augmentation strategy employing Poisson blending, inspired by the characteristics of X-ray images, to generate realistic synthetic instances of rare items which can effectively mitigate data imbalance; and (2) a context-aware feature enhancement module that captures the spatial and semantic interactions between objects and their surroundings, enhancing classification accuracy for underrepresented categories. Extensive experimental results demonstrate that DBNet improves detection performance for tail categories, outperforming sota methods in X-ray security inspection scenarios by a large margin 17.2%, thereby ensuring enhanced public safety., Comment: 10 pages, 4 figures

In contrast to conventional RIS, the scattering matrix of a non-reciprocal RIS (NR-RIS) is non-symmetric, leading to differences in the uplink and the downlink components of NR-RIS cascaded channels. In this paper, a physically-consistent device model is proposed in which an NR-RIS is composed of multiple groups of two-port elements inter-connected by non-reciprocal devices. The resulting non-reciprocal scattering matrix is derived for various cases including two-element groups connected with isolators or gyrators, and general three-element groups connected via circulators. Signal models are given for NR-RIS operating in either reflecting-only or simultaneously transmitting and reflecting modes. The problem of NR-RIS design for non-reciprocal beamsteering is formulated for three-element circulator implementations, and numerical results confirm that non-reciprocal beamsteering can be achieved with minimal sidelobe power. We also show that our physically consistent NR-RIS architecture is effective in implementing channel reciprocity attacks, achieving similar performance to that with idealized NR-RIS models., Comment: 13 Pages single Column

Computed Tomography (CT) is one of the most popular modalities for medical imaging. By far, CT images have contributed to the largest publicly available datasets for volumetric medical segmentation tasks, covering full-body anatomical structures. Large amounts of full-body CT images provide the opportunity to pre-train powerful models, e.g., STU-Net pre-trained in a supervised fashion, to segment numerous anatomical structures. However, it remains unclear in which conditions these pre-trained models can be transferred to various downstream medical segmentation tasks, particularly segmenting the other modalities and diverse targets. To address this problem, a large-scale benchmark for comprehensive evaluation is crucial for finding these conditions. Thus, we collected 87 public datasets varying in modality, target, and sample size to evaluate the transfer ability of full-body CT pre-trained models. We then employed a representative model, STU-Net with multiple model scales, to conduct transfer learning across modalities and targets. Our experimental results show that (1) there may be a bottleneck effect concerning the dataset size in fine-tuning, with more improvement on both small- and large-scale datasets than medium-size ones. (2) Models pre-trained on full-body CT demonstrate effective modality transfer, adapting well to other modalities such as MRI. (3) Pre-training on the full-body CT not only supports strong performance in structure detection but also shows efficacy in lesion detection, showcasing adaptability across target tasks. We hope that this large-scale open evaluation of transfer learning can direct future research in volumetric medical image segmentation.

Interactive Medical Image Segmentation (IMIS) has long been constrained by the limited availability of large-scale, diverse, and densely annotated datasets, which hinders model generalization and consistent evaluation across different models. In this paper, we introduce the IMed-361M benchmark dataset, a significant advancement in general IMIS research. First, we collect and standardize over 6.4 million medical images and their corresponding ground truth masks from multiple data sources. Then, leveraging the strong object recognition capabilities of a vision foundational model, we automatically generated dense interactive masks for each image and ensured their quality through rigorous quality control and granularity management. Unlike previous datasets, which are limited by specific modalities or sparse annotations, IMed-361M spans 14 modalities and 204 segmentation targets, totaling 361 million masks-an average of 56 masks per image. Finally, we developed an IMIS baseline network on this dataset that supports high-quality mask generation through interactive inputs, including clicks, bounding boxes, text prompts, and their combinations. We evaluate its performance on medical image segmentation tasks from multiple perspectives, demonstrating superior accuracy and scalability compared to existing interactive segmentation models. To facilitate research on foundational models in medical computer vision, we release the IMed-361M and model at https://github.com/uni-medical/IMIS-Bench.

In recent years, the programming capabilities of large language models (LLMs) have garnered significant attention. Fuzz testing, a highly effective technique, plays a key role in enhancing software reliability and detecting vulnerabilities. However, traditional fuzz testing tools rely on manually crafted fuzz drivers, which can limit both testing efficiency and effectiveness. To address this challenge, we propose an automated fuzz testing method driven by a code knowledge graph and powered by an LLM-based intelligent agent system, referred to as CKGFuzzer. We approach fuzz driver creation as a code generation task, leveraging the knowledge graph of the code repository to automate the generation process within the fuzzing loop, while continuously refining both the fuzz driver and input seeds. The code knowledge graph is constructed through interprocedural program analysis, where each node in the graph represents a code entity, such as a function or a file. The knowledge graph-enhanced CKGFuzzer not only effectively resolves compilation errors in fuzz drivers and generates input seeds tailored to specific API usage scenarios, but also analyzes fuzz driver crash reports, assisting developers in improving code quality. By querying the knowledge graph of the code repository and learning from API usage scenarios, we can better identify testing targets and understand the specific purpose of each fuzz driver. We evaluated our approach using eight open-source software projects. The experimental results indicate that CKGFuzzer achieved an average improvement of 8.73% in code coverage compared to state-of-the-art techniques. Additionally, CKGFuzzer reduced the manual review workload in crash case analysis by 84.4% and successfully detected 11 real bugs (including nine previously unreported bugs) across the tested libraries., Comment: 12 pages, 3 figures

Impersonation tactics, such as app squatting and app cloning, have posed longstanding challenges in mobile app stores, where malicious actors exploit the names and reputations of popular apps to deceive users. With the rapid growth of Large Language Model (LLM) stores like GPT Store and FlowGPT, these issues have similarly surfaced, threatening the integrity of the LLM app ecosystem. In this study, we present the first large-scale analysis of LLM app squatting and cloning using our custom-built tool, LLMappCrazy. LLMappCrazy covers 14 squatting generation techniques and integrates Levenshtein distance and BERT-based semantic analysis to detect cloning by analyzing app functional similarities. Using this tool, we generated variations of the top 1000 app names and found over 5,000 squatting apps in the dataset. Additionally, we observed 3,509 squatting apps and 9,575 cloning cases across six major platforms. After sampling, we find that 18.7% of the squatting apps and 4.9% of the cloning apps exhibited malicious behavior, including phishing, malware distribution, fake content dissemination, and aggressive ad injection.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Vision-language pre-training (VLP) models, trained on large-scale image-text pairs, have become widely used across a variety of downstream vision-and-language (V+L) tasks. This widespread adoption raises concerns about their vulnerability to adversarial attacks. Non-universal adversarial attacks, while effective, are often impractical for real-time online applications due to their high computational demands per data instance. Recently, universal adversarial perturbations (UAPs) have been introduced as a solution, but existing generator-based UAP methods are significantly time-consuming. To overcome the limitation, we propose a direct optimization-based UAP approach, termed DO-UAP, which significantly reduces resource consumption while maintaining high attack performance. Specifically, we explore the necessity of multimodal loss design and introduce a useful data augmentation strategy. Extensive experiments conducted on three benchmark VLP datasets, six popular VLP models, and three classical downstream tasks demonstrate the efficiency and effectiveness of DO-UAP. Specifically, our approach drastically decreases the time consumption by 23-fold while achieving a better attack performance., Comment: 11 pages

Accurate underwater target localization is essential for underwater exploration. To improve accuracy and efficiency in complex underwater environments, we propose the Electric Field Inversion-Localization Network (EFILN), a deep feedforward neural network that reconstructs position coordinates from underwater electric field signals. By assessing whether the neural network's input-output values satisfy the Coulomb law, the error between the network's inversion solution and the equation's exact solution can be determined. The Adam optimizer was employed first, followed by the L-BFGS optimizer, to progressively improve the output precision of EFILN. A series of noise experiments demonstrated the robustness and practical utility of the proposed method, while small sample data experiments validated its strong small-sample learning (SSL) capabilities. To accelerate relevant research, we have made the codes available as open-source.

Enlarging the context window of large language models (LLMs) has become a crucial research area, particularly for applications involving extremely long texts. In this work, we propose a novel training-free framework for processing long texts, utilizing a divide-and-conquer strategy to achieve comprehensive document understanding. The proposed LLM$\times$MapReduce framework splits the entire document into several chunks for LLMs to read and then aggregates the intermediate answers to produce the final output. The main challenge for divide-and-conquer long text processing frameworks lies in the risk of losing essential long-range information when splitting the document, which can lead the model to produce incomplete or incorrect answers based on the segmented texts. Disrupted long-range information can be classified into two categories: inter-chunk dependency and inter-chunk conflict. We design a structured information protocol to better cope with inter-chunk dependency and an in-context confidence calibration mechanism to resolve inter-chunk conflicts. Experimental results demonstrate that LLM$\times$MapReduce can outperform representative open-source and commercial long-context LLMs, and is applicable to several different models., Comment: Work in Progress. Code: https://github.com/thunlp/LLMxMapReduce

Code Large Language Models (LLMs) have demonstrated remarkable capabilities in generating, understanding, and manipulating programming code. However, their training process inadvertently leads to the memorization of sensitive information, posing severe privacy risks. Existing studies on memorization in LLMs primarily rely on prompt engineering techniques, which suffer from limitations such as widespread hallucination and inefficient extraction of the target sensitive information. In this paper, we present a novel approach to characterize real and fake secrets generated by Code LLMs based on token probabilities. We identify four key characteristics that differentiate genuine secrets from hallucinated ones, providing insights into distinguishing real and fake secrets. To overcome the limitations of existing works, we propose DESEC, a two-stage method that leverages token-level features derived from the identified characteristics to guide the token decoding process. DESEC consists of constructing an offline token scoring model using a proxy Code LLM and employing the scoring model to guide the decoding process by reassigning token likelihoods. Through extensive experiments on four state-of-the-art Code LLMs using a diverse dataset, we demonstrate the superior performance of DESEC in achieving a higher plausible rate and extracting more real secrets compared to existing baselines. Our findings highlight the effectiveness of our token-level approach in enabling an extensive assessment of the privacy leakage risks associated with Code LLMs., Comment: 12 pages, 8 figures

To keep pace with the rapid advancements in design complexity within modern computing systems, directed graph representation learning (DGRL) has become crucial, particularly for encoding circuit netlists, computational graphs, and developing surrogate models for hardware performance prediction. However, DGRL remains relatively unexplored, especially in the hardware domain, mainly due to the lack of comprehensive and user-friendly benchmarks. This study presents a novel benchmark comprising five hardware design datasets and 13 prediction tasks spanning various levels of circuit abstraction. We evaluate 21 DGRL models, employing diverse graph neural networks and graph transformers (GTs) as backbones, enhanced by positional encodings (PEs) tailored for directed graphs. Our results highlight that bidirected (BI) message passing neural networks (MPNNs) and robust PEs significantly enhance model performance. Notably, the top-performing models include PE-enhanced GTs interleaved with BI-MPNN layers and BI-Graph Isomorphism Network, both surpassing baselines across the 13 tasks. Additionally, our investigation into out-of-distribution (OOD) performance emphasizes the urgent need to improve OOD generalization in DGRL models. This benchmark, implemented with a modular codebase, streamlines the evaluation of DGRL models for both hardware and ML practitioners

Fine-tuning pre-trained models is crucial for adapting large models to downstream tasks, often delivering state-of-the-art performance. However, fine-tuning all model parameters is resource-intensive and laborious, leading to the emergence of parameter-efficient fine-tuning (PEFT) methods. One widely adopted PEFT technique, Low-Rank Adaptation (LoRA), freezes the pre-trained model weights and introduces two low-rank matrices whose ranks are significantly smaller than the dimensions of the original weight matrices. This enables efficient fine-tuning by adjusting only a small number of parameters. Despite its efficiency, LoRA approximates weight updates using low-rank decomposition, which struggles to capture complex, non-linear components and efficient optimization trajectories. As a result, LoRA-based methods often exhibit a significant performance gap compared to full fine-tuning. Closing this gap requires higher ranks, which increases the number of parameters. To address these limitations, we propose a nonlinear parameter-efficient adaptation method (NEAT). NEAT introduces a lightweight neural network that takes pre-trained weights as input and learns a nonlinear transformation to approximate cumulative weight updates. These updates can be interpreted as functions of the corresponding pre-trained weights. The nonlinear approximation directly models the cumulative updates, effectively capturing complex and non-linear structures in the weight updates. Our theoretical analysis demonstrates taht NEAT can be more efficient than LoRA while having equal or greater expressivity. Extensive evaluations across four benchmarks and over twenty datasets demonstrate that NEAT significantly outperforms baselines in both vision and text tasks.

We present the Melody-Guided Music Generation (MG2) model, a novel approach using melody to guide the text-to-music generation that, despite a simple method and limited resources, achieves excellent performance. Specifically, we first align the text with audio waveforms and their associated melodies using the newly proposed Contrastive Language-Music Pretraining, enabling the learned text representation fused with implicit melody information. Subsequently, we condition the retrieval-augmented diffusion module on both text prompt and retrieved melody. This allows MG2 to generate music that reflects the content of the given text description, meantime keeping the intrinsic harmony under the guidance of explicit melody information. We conducted extensive experiments on two public datasets: MusicCaps and MusicBench. Surprisingly, the experimental results demonstrate that the proposed MG2 model surpasses current open-source text-to-music generation models, achieving this with fewer than 1/3 of the parameters or less than 1/200 of the training data compared to state-of-the-art counterparts. Furthermore, we conducted comprehensive human evaluations involving three types of users and five perspectives, using newly designed questionnaires to explore the potential real-world applications of MG2., Comment: 16 pages, 8 figure, 8 tables

Periodic soft real-time systems have broad applications in many areas, such as IoT. Finding an optimal energy-efficient policy that is adaptable to underlying edge devices while meeting deadlines for tasks has always been challenging. This research studies generalized systems with multi-task, multi-deadline scenarios with reinforcement learning-based DVFS for energy saving. This work addresses the limitation of previous work that models a periodic system as a single task and single-deadline scenario, which is too simplified to cope with complex situations. The method encodes time series information in the Linux kernel into information that is easy to use for reinforcement learning, allowing the system to generate DVFS policies to adapt system patterns based on the general workload. For encoding, we present two different methods for comparison. Both methods use only one performance counter: system utilization and the kernel only needs minimal information from the userspace. Our method is implemented on Jetson Nano Board (2GB) and is tested with three fixed multitask workloads, which are three, five, and eight tasks in the workload, respectively. For randomness and generalization, we also designed a random workload generator to build different multitask workloads to test. Based on the test results, our method could save 3%-10% power compared to Linux built-in governors.

Recent researches have proven that pre-training on large-scale person images extracted from internet videos is an effective way in learning better representations for person re-identification. However, these researches are mostly confined to pre-training at the instance-level or single-video tracklet-level. They ignore the identity-invariance in images of the same person across different videos, which is a key focus in person re-identification. To address this issue, we propose a Cross-video Identity-cOrrelating pre-traiNing (CION) framework. Defining a noise concept that comprehensively considers both intra-identity consistency and inter-identity discrimination, CION seeks the identity correlation from cross-video images by modeling it as a progressive multi-level denoising problem. Furthermore, an identity-guided self-distillation loss is proposed to implement better large-scale pre-training by mining the identity-invariance within person images. We conduct extensive experiments to verify the superiority of our CION in terms of efficiency and performance. CION achieves significantly leading performance with even fewer training samples. For example, compared with the previous state-of-the-art~\cite{ISR}, CION with the same ResNet50-IBN achieves higher mAP of 93.3\% and 74.3\% on Market1501 and MSMT17, while only utilizing 8\% training samples. Finally, with CION demonstrating superior model-agnostic ability, we contribute a model zoo named ReIDZoo to meet diverse research and application needs in this field. It contains a series of CION pre-trained models with spanning structures and parameters, totaling 32 models with 10 different structures, including GhostNet, ConvNext, RepViT, FastViT and so on. The code and models will be made publicly available at https://github.com/Zplusdragon/CION_ReIDZoo., Comment: NeurIPS 2024 Accepted Paper

Recent advancements in speech synthesis models, trained on extensive datasets, have demonstrated remarkable zero-shot capabilities. These models can control content, timbre, and emotion in generated speech based on prompt inputs. Despite these advancements, the choice of prompts significantly impacts the output quality, yet most existing selection schemes do not adequately address the control of emotional intensity. To address this question, this paper proposes a two-stage prompt selection strategy EmoPro, which is specifically designed for emotionally controllable speech synthesis. This strategy focuses on selecting highly expressive and high-quality prompts by evaluating them from four perspectives: emotional expression strength, speech quality, text-emotion consistency, and model generation performance. Experimental results show that prompts selected using the proposed method result in more emotionally expressive and engaging synthesized speech compared to those obtained through baseline. Audio samples and codes will be available at https://whyrrrrun.github.io/EmoPro/.

In recent years, Large Language Models (LLMs) have gained widespread use, raising concerns about their security. Traditional jailbreak attacks, which often rely on the model internal information or have limitations when exploring the unsafe behavior of the victim model, limiting their reducing their general applicability. In this paper, we introduce PathSeeker, a novel black-box jailbreak method, which is inspired by the game of rats escaping a maze. We think that each LLM has its unique "security maze", and attackers attempt to find the exit learning from the received feedback and their accumulated experience to compromise the target LLM's security defences. Our approach leverages multi-agent reinforcement learning, where smaller models collaborate to guide the main LLM in performing mutation operations to achieve the attack objectives. By progressively modifying inputs based on the model's feedback, our system induces richer, harmful responses. During our manual attempts to perform jailbreak attacks, we found that the vocabulary of the response of the target model gradually became richer and eventually produced harmful responses. Based on the observation, we also introduce a reward mechanism that exploits the expansion of vocabulary richness in LLM responses to weaken security constraints. Our method outperforms five state-of-the-art attack techniques when tested across 13 commercial and open-source LLMs, achieving high attack success rates, especially in strongly aligned commercial models like GPT-4o-mini, Claude-3.5, and GLM-4-air with strong safety alignment. This study aims to improve the understanding of LLM security vulnerabilities and we hope that this sturdy can contribute to the development of more robust defenses., Comment: update the abstract and cite a new related work

Declarative UI frameworks have gained widespread adoption in mobile app development, offering benefits such as improved code readability and easier maintenance. Despite these advantages, the process of translating UI designs into functional code remains challenging and time-consuming. Recent advancements in multimodal large language models (MLLMs) have shown promise in directly generating mobile app code from user interface (UI) designs. However, the direct application of MLLMs to this task is limited by challenges in accurately recognizing UI components and comprehensively capturing interaction logic. To address these challenges, we propose DeclarUI, an automated approach that synergizes computer vision (CV), MLLMs, and iterative compiler-driven optimization to generate and refine declarative UI code from designs. DeclarUI enhances visual fidelity, functional completeness, and code quality through precise component segmentation, Page Transition Graphs (PTGs) for modeling complex inter-page relationships, and iterative optimization. In our evaluation, DeclarUI outperforms baselines on React Native, a widely adopted declarative UI framework, achieving a 96.8% PTG coverage rate and a 98% compilation success rate. Notably, DeclarUI demonstrates significant improvements over state-of-the-art MLLMs, with a 123% increase in PTG coverage rate, up to 55% enhancement in visual similarity scores, and a 29% boost in compilation success rate. We further demonstrate DeclarUI's generalizability through successful applications to Flutter and ArkUI frameworks.

The proliferation of pre-trained models (PTMs) and datasets has led to the emergence of centralized model hubs like Hugging Face, which facilitate collaborative development and reuse. However, recent security reports have uncovered vulnerabilities and instances of malicious attacks within these platforms, highlighting growing security concerns. This paper presents the first systematic study of malicious code poisoning attacks on pre-trained model hubs, focusing on the Hugging Face platform. We conduct a comprehensive threat analysis, develop a taxonomy of model formats, and perform root cause analysis of vulnerable formats. While existing tools like Fickling and ModelScan offer some protection, they face limitations in semantic-level analysis and comprehensive threat detection. To address these challenges, we propose MalHug, an end-to-end pipeline tailored for Hugging Face that combines dataset loading script extraction, model deserialization, in-depth taint analysis, and heuristic pattern matching to detect and classify malicious code poisoning attacks in datasets and models. In collaboration with Ant Group, a leading financial technology company, we have implemented and deployed MalHug on a mirrored Hugging Face instance within their infrastructure, where it has been operational for over three months. During this period, MalHug has monitored more than 705K models and 176K datasets, uncovering 91 malicious models and 9 malicious dataset loading scripts. These findings reveal a range of security threats, including reverse shell, browser credential theft, and system reconnaissance. This work not only bridges a critical gap in understanding the security of the PTM supply chain but also provides a practical, industry-tested solution for enhancing the security of pre-trained model hubs., Comment: To appear in the 39th IEEE/ACM International Conference on Automated Software Engineering (ASE'24), October 27-November 1, 2024, Sacramento, CA, USA

The exponential growth of open-source package ecosystems, particularly NPM and PyPI, has led to an alarming increase in software supply chain poisoning attacks. Existing static analysis methods struggle with high false positive rates and are easily thwarted by obfuscation and dynamic code execution techniques. While dynamic analysis approaches offer improvements, they often suffer from capturing non-package behaviors and employing simplistic testing strategies that fail to trigger sophisticated malicious behaviors. To address these challenges, we present OSCAR, a robust dynamic code poisoning detection pipeline for NPM and PyPI ecosystems. OSCAR fully executes packages in a sandbox environment, employs fuzz testing on exported functions and classes, and implements aspect-based behavior monitoring with tailored API hook points. We evaluate OSCAR against six existing tools using a comprehensive benchmark dataset of real-world malicious and benign packages. OSCAR achieves an F1 score of 0.95 in NPM and 0.91 in PyPI, confirming that OSCAR is as effective as the current state-of-the-art technologies. Furthermore, for benign packages exhibiting characteristics typical of malicious packages, OSCAR reduces the false positive rate by an average of 32.06% in NPM (from 34.63% to 2.57%) and 39.87% in PyPI (from 41.10% to 1.23%), compared to other tools, significantly reducing the workload of manual reviews in real-world deployments. In cooperation with Ant Group, a leading financial technology company, we have deployed OSCAR on its NPM and PyPI mirrors since January 2023, identifying 10,404 malicious NPM packages and 1,235 malicious PyPI packages over 18 months. This work not only bridges the gap between academic research and industrial application in code poisoning detection but also provides a robust and practical solution that has been thoroughly tested in a real-world industrial setting., Comment: To appear in the 39th IEEE/ACM International Conference on Automated Software Engineering (ASE'24 Industry Showcase), October 27-November 1, 2024, Sacramento, CA, USA

With the rapid advancement of technologies like text-to-speech (TTS) and voice conversion (VC), detecting deepfake voices has become increasingly crucial. However, both academia and industry lack a comprehensive and intuitive benchmark for evaluating detectors. Existing datasets are limited in language diversity and lack many manipulations encountered in real-world production environments. To fill this gap, we propose VoiceWukong, a benchmark designed to evaluate the performance of deepfake voice detectors. To build the dataset, we first collected deepfake voices generated by 19 advanced and widely recognized commercial tools and 15 open-source tools. We then created 38 data variants covering six types of manipulations, constructing the evaluation dataset for deepfake voice detection. VoiceWukong thus includes 265,200 English and 148,200 Chinese deepfake voice samples. Using VoiceWukong, we evaluated 12 state-of-the-art detectors. AASIST2 achieved the best equal error rate (EER) of 13.50%, while all others exceeded 20%. Our findings reveal that these detectors face significant challenges in real-world applications, with dramatically declining performance. In addition, we conducted a user study with more than 300 participants. The results are compared with the performance of the 12 detectors and a multimodel large language model (MLLM), i.e., Qwen2-Audio, where different detectors and humans exhibit varying identification capabilities for deepfake voices at different deception levels, while the LALM demonstrates no detection ability at all. Furthermore, we provide a leaderboard for deepfake voice detection, publicly available at {https://voicewukong.github.io}.

Ransomware attacks have emerged as one of the most significant cybersecurity threats. Despite numerous proposed detection and defense methods, existing approaches face two fundamental limitations in large-scale industrial applications: intolerable system overheads and notorious alert fatigue. To address these challenges, we propose CanCal, a real-time and lightweight ransomware detection system. Specifically, CanCal selectively filters suspicious processes by the monitoring layers and then performs in-depth behavioral analysis to isolate ransomware activities from benign operations, minimizing alert fatigue while ensuring lightweight computational and storage overhead. The experimental results on a large-scale industrial environment~(1,761 ransomware, ~3 million events, continuous test over 5 months) indicate that CanCal is as effective as state-of-the-art techniques while enabling rapid inference within 30ms and real-time response within a maximum of 3 seconds. CanCal dramatically reduces average CPU utilization by 91.04% (from 6.7% to 0.6%) and peak CPU utilization by 76.69% (from 26.6% to 6.2%), while avoiding 76.50% (from 3,192 to 750) of the inspection efforts from security analysts. By the time of this writing, CanCal has been integrated into a commercial product and successfully deployed on 3.32 million endpoints for over a year. From March 2023 to April 2024, CanCal successfully detected and thwarted 61 ransomware attacks, demonstrating the effectiveness of CanCal in combating sophisticated ransomware threats in real-world scenarios., Comment: To appear in the 2024 ACM SIGSAC Conference on Computer and Communications Security (CCS'24), October 14--18, 2024, Salt Lake City

The swift advancement of large language models (LLMs) has profoundly shaped the landscape of artificial intelligence; however, their deployment in sensitive domains raises grave concerns, particularly due to their susceptibility to malicious exploitation. This situation underscores the insufficiencies in pre-deployment testing, highlighting the urgent need for more rigorous and comprehensive evaluation methods. This study presents a comprehensive empirical analysis assessing the efficacy of conventional coverage criteria in identifying these vulnerabilities, with a particular emphasis on the pressing issue of jailbreak attacks. Our investigation begins with a clustering analysis of the hidden states in LLMs, demonstrating that intrinsic characteristics of these states can distinctly differentiate between various types of queries. Subsequently, we assess the performance of these criteria across three critical dimensions: criterion level, layer level, and token level. Our findings uncover significant disparities in neuron activation patterns between the processing of normal and jailbreak queries, thereby corroborating the clustering results. Leveraging these findings, we propose an innovative approach for the real-time detection of jailbreak attacks by utilizing neural activation features. Our classifier demonstrates remarkable accuracy, averaging 96.33% in identifying jailbreak queries, including those that could lead to adversarial attacks. The importance of our research lies in its comprehensive approach to addressing the intricate challenges of LLM security. By enabling instantaneous detection from the model's first token output, our method holds promise for future systems integrating LLMs, offering robust real-time detection capabilities. This study advances our understanding of LLM security testing, and lays a critical foundation for the development of more resilient AI systems.

As the most powerful tool in discrepancy theory, the partial coloring method has wide applications in many problems including the Beck-Fiala problem and Spencer's celebrated result. Currently, there are two major algorithmic methods for the partial coloring method: the first approach uses linear algebraic tools; and the second is called Gaussian measure algorithm. We explore the advantages of these two methods and show the following results for them separately. 1. Spencer conjectured that the prefix discrepancy of any $\mathbf{A} \in \{0,1\}^{m \times n}$ is $O(\sqrt{m})$. We show how to find a partial coloring with prefix discrepancy $O(\sqrt{m})$ and $\Omega(n)$ entries in $\{ \pm 1\}$ efficiently. To the best of our knowledge, this provides the first partial coloring whose prefix discrepancy is almost optimal. However, unlike the classical discrepancy problem, there is no reduction on the number of variables $n$ for the prefix problem. By recursively applying partial coloring, we obtain a full coloring with prefix discrepancy $O(\sqrt{m} \cdot \log \frac{O(n)}{m})$. Prior to this work, the best bounds of the prefix Spencer conjecture for arbitrarily large $n$ were $2m$ and $O(\sqrt{m \log n})$. 2. Our second result extends the first linear algebraic approach to a sampling algorithm in Spencer's classical setting. On the first hand, Spencer proved that there are $1.99^m$ good colorings with discrepancy $O(\sqrt{m})$. Hence a natural question is to design efficient random sampling algorithms in Spencer's setting. On the other hand, some applications of discrepancy theory, prefer a random solution instead of a fixed one. Our second result is an efficient sampling algorithm whose random output has min-entropy $\Omega(n)$ and discrepancy $O(\sqrt{m})$. Moreover, our technique extends the linear algebraic framework by incorporating leverage scores of randomized matrix algorithms.

Large-scale data collection is essential for developing personalized training data, mitigating the shortage of training data, and fine-tuning specialized models. However, creating high-quality datasets quickly and accurately remains a challenge due to annotation errors, the substantial time and costs associated with human labor. To address these issues, we propose Automatic Dataset Construction (ADC), an innovative methodology that automates dataset creation with negligible cost and high efficiency. Taking the image classification task as a starting point, ADC leverages LLMs for the detailed class design and code generation to collect relevant samples via search engines, significantly reducing the need for manual annotation and speeding up the data generation process. Despite these advantages, ADC also encounters real-world challenges such as label errors (label noise) and imbalanced data distributions (label bias). We provide open-source software that incorporates existing methods for label error detection, robust learning under noisy and biased data, ensuring a higher-quality training data and more robust model training procedure. Furthermore, we design three benchmark datasets focused on label noise detection, label noise learning, and class-imbalanced learning. These datasets are vital because there are few existing datasets specifically for label noise detection, despite its importance. Finally, we evaluate the performance of existing popular methods on these datasets, thereby facilitating further research in the field.

Large Language Models (LLMs) are implicit troublemakers. While they provide valuable insights and assist in problem-solving, they can also potentially serve as a resource for malicious activities. Implementing safety alignment could mitigate the risk of LLMs generating harmful responses. We argue that: even when an LLM appears to successfully block harmful queries, there may still be hidden vulnerabilities that could act as ticking time bombs. To identify these underlying weaknesses, we propose to use a cost value model as both a detector and an attacker. Trained on external or self-generated harmful datasets, the cost value model could successfully influence the original safe LLM to output toxic content in decoding process. For instance, LLaMA-2-chat 7B outputs 39.18% concrete toxic content, along with only 22.16% refusals without any harmful suffixes. These potential weaknesses can then be exploited via prompt optimization such as soft prompts on images. We name this decoding strategy: Jailbreak Value Decoding (JVD), emphasizing that seemingly secure LLMs may not be as safe as we initially believe. They could be used to gather harmful data or launch covert attacks.

WebAssembly (Wasm), as a compact, fast, and isolation-guaranteed binary format, can be compiled from more than 40 high-level programming languages. However, vulnerabilities in Wasm binaries could lead to sensitive data leakage and even threaten their hosting environments. To identify them, symbolic execution is widely adopted due to its soundness and the ability to automatically generate exploitations. However, existing symbolic executors for Wasm binaries are typically platform-specific, which means that they cannot support all Wasm features. They may also require significant manual interventions to complete the analysis and suffer from efficiency issues as well. In this paper, we propose an efficient and fully-functional symbolic execution engine, named SeeWasm. Compared with existing tools, we demonstrate that SeeWasm supports full-featured Wasm binaries without further manual intervention, while accelerating the analysis by 2 to 6 times. SeeWasm has been adopted by existing works to identify more than 30 0-day vulnerabilities or security issues in well-known C, Go, and SGX applications after compiling them to Wasm binaries., Comment: Accepted by ISSTA'24 Demo Track, the tool can be accessed at https://github.com/PKU-ASAL/SeeWasm

Large language models (LLMs) have achieved unprecedented success in the field of natural language processing. However, the black-box nature of their internal mechanisms has brought many concerns about their trustworthiness and interpretability. Recent research has discovered a class of abnormal tokens in the model's vocabulary space and named them "glitch tokens". Those tokens, once included in the input, may induce the model to produce incorrect, irrelevant, or even harmful results, drastically undermining the reliability and practicality of LLMs. In this work, we aim to enhance the understanding of glitch tokens and propose techniques for their detection and mitigation. We first reveal the characteristic features induced by glitch tokens on LLMs, which are evidenced by significant deviations in the distributions of attention patterns and dynamic information from intermediate model layers. Based on the insights, we develop GlitchProber, a tool for efficient glitch token detection and mitigation. GlitchProber utilizes small-scale sampling, principal component analysis for accelerated feature extraction, and a simple classifier for efficient vocabulary screening. Taking one step further, GlitchProber rectifies abnormal model intermediate layer values to mitigate the destructive effects of glitch tokens. Evaluated on five mainstream open-source LLMs, GlitchProber demonstrates higher efficiency, precision, and recall compared to existing approaches, with an average F1 score of 0.86 and an average repair rate of 50.06%. GlitchProber unveils a novel path to address the challenges posed by glitch tokens and inspires future research toward more robust and interpretable LLMs.

In contemporary software development, the widespread use of indirect calls to achieve dynamic features poses challenges in constructing precise control flow graphs (CFGs), which further impacts the performance of downstream static analysis tasks. To tackle this issue, various types of indirect call analyzers have been proposed. However, they do not fully leverage the semantic information of the program, limiting their effectiveness in real-world scenarios. To address these issues, this paper proposes Semantic-Enhanced Analysis (SEA), a new approach to enhance the effectiveness of indirect call analysis. Our fundamental insight is that for common programming practices, indirect calls often exhibit semantic similarity with their invoked targets. This semantic alignment serves as a supportive mechanism for static analysis techniques in filtering out false targets. Notably, contemporary large language models (LLMs) are trained on extensive code corpora, encompassing tasks such as code summarization, making them well-suited for semantic analysis. Specifically, SEA leverages LLMs to generate natural language summaries of both indirect calls and target functions from multiple perspectives. Through further analysis of these summaries, SEA can determine their suitability as caller-callee pairs. Experimental results demonstrate that SEA can significantly enhance existing static analysis methods by producing more precise target sets for indirect calls., Comment: Accepted by ASE'24

Within the thriving mobile app ecosystem ecosystem, a subset of apps provides illicit services such as gambling and pornography to pursue economic gains, collectively referred to as "underground economy apps". While previous studies have examined these apps' characteristics and identification methods, investigations into their distribution via platforms beyond app markets (like Telegram) remain scarce, which has emerged as a crucial channel for underground activities and cybercrime due to the robust encryption and user anonymity. This study provides the first comprehensive exploration of the underground mobile app ecosystem on Telegram. Overcoming the complexities of the Telegram environment, we build a novel dataset and analyze the prevalence, promotional strategies, and characteristics of these apps. Our findings reveal the significant prevalence of these apps on Telegram, with the total sum of subscription user numbers across channels promoting these apps equivalent to 1% of Telegram's user base. We find these apps primarily cater to gambling and pornography services. We uncover sophisticated promotional strategies involving complex networks of apps, websites, users, and channels, and identify significant gaps in Telegram's content moderation capabilities. Our analysis also exposes the misuse of iOS features for app distribution and the prevalence of malicious behaviors in these apps. This research not only enhances our understanding of the underground app ecosystem but also provides valuable insights for developing effective regulatory measures and protecting users from potential risks associated with these covert operations. Our findings provide implications for platform regulators, app market operators, law enforcement agencies, and cybersecurity professionals in combating the proliferation of underground apps on encrypted messaging platforms., Comment: To appear in SIGMETRICS 2025

OpenAI's advanced large language models (LLMs) have revolutionized natural language processing and enabled developers to create innovative applications. As adoption grows, understanding the experiences and challenges of developers working with these technologies is crucial. This paper presents a comprehensive analysis of the OpenAI Developer Forum, focusing on (1) popularity trends and user engagement patterns, and (2) a taxonomy of challenges and concerns faced by developers. We first employ a quantitative analysis of the metadata from 29,576 forum topics, investigating temporal trends in topic creation, the popularity of topics across different categories, and user contributions at various trust levels. We then qualitatively analyze content from 9,301 recently active topics on developer concerns. From a sample of 886 topics, we construct a taxonomy of concerns in the OpenAI Developer Forum. Our findings uncover critical concerns raised by developers in creating AI-powered applications and offer targeted recommendations to address them. This work not only advances AI-assisted software engineering but also empowers developer communities to shape the responsible evolution and integration of AI technology in society.