Min Yang

Mi Zhang, Xudong Pan, Min Yang

In this paper, we present JADE, a targeted linguistic fuzzing platform which strengthens the linguistic complexity of seed questions to simultaneously and consistently break a wide range of widely-used LLMs categorized in three groups: eight open-sourced Chinese, six commercial Chinese and four commercial English LLMs. JADE generates three safety benchmarks for the three groups of LLMs, which contain unsafe questions that are highly threatening: the questions simultaneously trigger harmful generation of multiple LLMs, with an average unsafe generation ratio of $70\%$ (please see the table below), while are still natural questions, fluent and preserving the core unsafe semantics. We release the benchmark demos generated for commercial English LLMs and open-sourced English LLMs in the following link: https://github.com/whitzard-ai/jade-db. For readers who are interested in evaluating on more questions generated by JADE, please contact us. JADE is based on Noam Chomsky's seminal theory of transformational-generative grammar. Given a seed question with unsafe intention, JADE invokes a sequence of generative and transformational rules to increment the complexity of the syntactic structure of the original question, until the safety guardrail is broken. Our key insight is: Due to the complexity of human language, most of the current best LLMs can hardly recognize the invariant evil from the infinite number of different syntactic structures which form an unbound example space that can never be fully covered. Technically, the generative/transformative rules are constructed by native speakers of the languages, and, once developed, can be used to automatically grow and transform the parse tree of a given question, until the guardrail is broken. For more evaluation results and demo, please check our website: https://whitzard-ai.github.io/jade.html.

Yifan Lu, Wenxuan Li, Mi Zhang et al.

To protect the intellectual property of well-trained deep neural networks (DNNs), black-box watermarks, which are embedded into the prediction behavior of DNN models on a set of specially-crafted samples and extracted from suspect models using only API access, have gained increasing popularity in both academy and industry. Watermark robustness is usually implemented against attackers who steal the protected model and obfuscate its parameters for watermark removal. However, current robustness evaluations are primarily performed under moderate attacks or unrealistic settings. Existing removal attacks could only crack a small subset of the mainstream black-box watermarks, and fall short in four key aspects: incomplete removal, reliance on prior knowledge of the watermark, performance degradation, and high dependency on data. In this paper, we propose a watermark-agnostic removal attack called \textsc{Neural Dehydration} (\textit{abbrev.} \textsc{Dehydra}), which effectively erases all ten mainstream black-box watermarks from DNNs, with only limited or even no data dependence. In general, our attack pipeline exploits the internals of the protected model to recover and unlearn the watermark message. We further design target class detection and recovered sample splitting algorithms to reduce the utility loss and achieve data-free watermark removal on five of the watermarking schemes. We conduct comprehensive evaluation of \textsc{Dehydra} against ten mainstream black-box watermarks on three benchmark datasets and DNN architectures. Compared with existing removal attacks, \textsc{Dehydra} achieves strong removal effectiveness across all the covered watermarks, preserving at least $90\%$ of the stolen model utility, under the data-limited settings, i.e., less than $2\%$ of the training data or even data-free.

Ruisong Gao, Min Yang, Jin Zhang

We propose a new class of physics-informed neural networks, called Physics-Informed Generator-Encoder Adversarial Networks, to effectively address the challenges posed by forward, inverse, and mixed problems in stochastic differential equations. In these scenarios, while the governing equations are known, the available data consist of only a limited set of snapshots for system parameters. Our model consists of two key components: the generator and the encoder, both updated alternately by gradient descent. In contrast to previous approaches of directly matching the approximated solutions with real snapshots, we employ an indirect matching that operates within the lower-dimensional latent feature space. This method circumvents challenges associated with high-dimensional inputs and complex data distributions, while yielding more accurate solutions compared to existing neural network solvers. In addition, the approach also mitigates the training instability issues encountered in previous adversarial frameworks in an efficient manner. Numerical results provide compelling evidence of the effectiveness of the proposed method in solving different types of stochastic differential equations.

Fei Wang, Ruohui Zhang, Chenglin Chen et al.

Multi-Object Tracking (MOT) aims to maintain stable and uninterrupted trajectories for each target. Most state-of-the-art approaches first detect objects in each frame and then implement data association between new detections and existing tracks using motion models and appearance similarities. Despite achieving satisfactory results, occlusion and crowds can easily lead to missing and distorted detections, followed by missing and false associations. In this paper, we first revisit the classic tracker DeepSORT, enhancing its robustness over crowds and occlusion significantly by placing greater trust in predictions when detections are unavailable or of low quality in crowded and occluded scenes. Specifically, we propose a new framework comprising of three lightweight and plug-and-play algorithms: the probability map, the prediction map, and the covariance adaptive Kalman filter. The probability map identifies whether undetected objects have genuinely disappeared from view (e.g., out of the image or entered a building) or are only temporarily undetected due to occlusion or other reasons. Trajectories of undetected targets that are still within the probability map are extended by state estimations directly. The prediction map determines whether an object is in a crowd, and we prioritize state estimations over observations when severe deformation of observations occurs, accomplished through the covariance adaptive Kalman filter. The proposed method, named MapTrack, achieves state-of-the-art results on popular multi-object tracking benchmarks such as MOT17 and MOT20. Despite its superior performance, our method remains simple, online, and real-time. The code will be open-sourced later.

Jian Wang, Junhao Liu, Wei Bi et al.

Neural network models usually suffer from the challenge of incorporating commonsense knowledge into the open-domain dialogue systems. In this paper, we propose a novel knowledge-aware dialogue generation model (called TransDG), which transfers question representation and knowledge matching abilities from knowledge base question answering (KBQA) task to facilitate the utterance understanding and factual knowledge selection for dialogue generation. In addition, we propose a response guiding attention and a multi-step decoding strategy to steer our model to focus on relevant features for response generation. Experiments on two benchmark datasets demonstrate that our model has robust superiority over compared methods in generating informative and fluent dialogues. Our code is available at https://github.com/siat-nlp/TransDG.

Zhicheng Lu, Xiang Guo, Le Hui et al.

In this paper, we propose a 3D geometry-aware deformable Gaussian Splatting method for dynamic view synthesis. Existing neural radiance fields (NeRF) based solutions learn the deformation in an implicit manner, which cannot incorporate 3D scene geometry. Therefore, the learned deformation is not necessarily geometrically coherent, which results in unsatisfactory dynamic view synthesis and 3D dynamic reconstruction. Recently, 3D Gaussian Splatting provides a new representation of the 3D scene, building upon which the 3D geometry could be exploited in learning the complex 3D deformation. Specifically, the scenes are represented as a collection of 3D Gaussian, where each 3D Gaussian is optimized to move and rotate over time to model the deformation. To enforce the 3D scene geometry constraint during deformation, we explicitly extract 3D geometry features and integrate them in learning the 3D deformation. In this way, our solution achieves 3D geometry-aware deformation modeling, which enables improved dynamic view synthesis and 3D dynamic reconstruction. Extensive experimental results on both synthetic and real datasets prove the superiority of our solution, which achieves new state-of-the-art performance. The project is available at https://npucvr.github.io/GaGS/

Min Yang, Huan Gao, Ping Guo et al.

Vision Transformer (ViT) has shown high potential in video recognition, owing to its flexible design, adaptable self-attention mechanisms, and the efficacy of masked pre-training. Yet, it remains unclear how to adapt these pre-trained short-term ViTs for temporal action detection (TAD) in untrimmed videos. The existing works treat them as off-the-shelf feature extractors for each short-trimmed snippet without capturing the fine-grained relation among different snippets in a broader temporal context. To mitigate this issue, this paper focuses on designing a new mechanism for adapting these pre-trained ViT models as a unified long-form video transformer to fully unleash its modeling power in capturing inter-snippet relation, while still keeping low computation overhead and memory consumption for efficient TAD. To this end, we design effective cross-snippet propagation modules to gradually exchange short-term video information among different snippets from two levels. For inner-backbone information propagation, we introduce a cross-snippet propagation strategy to enable multi-snippet temporal feature interaction inside the backbone.For post-backbone information propagation, we propose temporal transformer layers for further clip-level modeling. With the plain ViT-B pre-trained with VideoMAE, our end-to-end temporal action detector (ViT-TAD) yields a very competitive performance to previous temporal action detectors, riching up to 69.5 average mAP on THUMOS14, 37.40 average mAP on ActivityNet-1.3 and 17.20 average mAP on FineAction.

Yining Wang, Yuquan Wang, Xi Li et al.

As Large Language Models (LLMs) continue to advance, they are increasingly relied upon as real-time sources of information by non-expert users. To ensure the factuality of the information they provide, much research has focused on mitigating hallucinations in LLM responses, but only in the context of formal user queries, rather than maliciously crafted ones. In this study, we introduce The Illusionist's Prompt, a novel hallucination attack that incorporates linguistic nuances into adversarial queries, challenging the factual accuracy of LLMs against five types of fact-enhancing strategies. Our attack automatically generates highly transferrable illusory prompts to induce internal factual errors, all while preserving user intent and semantics. Extensive experiments confirm the effectiveness of our attack in compromising black-box LLMs, including commercial APIs like GPT-4o and Gemini-2.0, even with various defensive mechanisms.

Run Luo, Zikai Song, Longze Chen et al.

Multi-Object Tracking (MOT) aims to associate multiple objects across video frames and is a challenging vision task due to inherent complexities in the tracking environment. Most existing approaches train and track within a single domain, resulting in a lack of cross-domain generalizability to data from other domains. While several works have introduced natural language representation to bridge the domain gap in visual tracking, these textual descriptions often provide too high-level a view and fail to distinguish various instances within the same class. In this paper, we address this limitation by developing IP-MOT, an end-to-end transformer model for MOT that operates without concrete textual descriptions. Our approach is underpinned by two key innovations: Firstly, leveraging a pre-trained vision-language model, we obtain instance-level pseudo textual descriptions via prompt-tuning, which are invariant across different tracking scenes; Secondly, we introduce a query-balanced strategy, augmented by knowledge distillation, to further boost the generalization capabilities of our model. Extensive experiments conducted on three widely used MOT benchmarks, including MOT17, MOT20, and DanceTrack, demonstrate that our approach not only achieves competitive performance on same-domain data compared to state-of-the-art models but also significantly improves the performance of query-based trackers by large margins for cross-domain inputs.

Shiwen Ni, Min Yang, Ruifeng Xu et al.

Among the various pre-trained neural language models that are popular today, dropout is already an indispensable regularization technique. To solve the inconsistency between training and inference caused by the randomness of dropout, some studies use consistency training to regularize dropout at the output layer. In this paper, we propose a novel Layer-wise Regularized Dropout (LR-Drop), which is specially designed for Transformer-based Language models. Specifically, LR-Drop layer-wise regularizes each Transformer layer using the consistency training strategy. Each training sample passes through the two siamese sub-models sampled by dropout, and then LR-Drop forces the hidden states, multi-head attention matrices, and output distribution of the two siamese sub-models to be consistent. The proposed LR-Drop can be regarded as a "self-distillation" framework, in which each sub-model generated by dropout is the other's "teacher" model and "student" model. Through extensive experiments on 8 natural language understanding datasets, 6 neural machine translation datasets, and 1 abstractive summarization dataset (a total of 15 datasets), we show that LR-Drop achieves superior performances, including state-of-the-art results.

Dingmin Wang, Ziyao Chen, Wanwei He et al.

Most existing neural network based task-oriented dialogue systems follow encoder-decoder paradigm, where the decoder purely depends on the source texts to generate a sequence of words, usually suffering from instability and poor readability. Inspired by the traditional template-based generation approaches, we propose a template-guided hybrid pointer network for the knowledge-based task-oriented dialogue system, which retrieves several potentially relevant answers from a pre-constructed domain-specific conversational repository as guidance answers, and incorporates the guidance answers into both the encoding and decoding processes. Specifically, we design a memory pointer network model with a gating mechanism to fully exploit the semantic correlation between the retrieved answers and the ground-truth response. We evaluate our model on four widely used task-oriented datasets, including one simulated and three manually created datasets. The experimental results demonstrate that the proposed model achieves significantly better performance than the state-of-the-art methods over different automatic evaluation metrics.

Haojie Pan, Junpei Zhou, Zhou Zhao et al.

We first propose a new task named Dialogue Description (Dial2Desc). Unlike other existing dialogue summarization tasks such as meeting summarization, we do not maintain the natural flow of a conversation but describe an object or an action of what people are talking about. The Dial2Desc system takes a dialogue text as input, then outputs a concise description of the object or the action involved in this conversation. After reading this short description, one can quickly extract the main topic of a conversation and build a clear picture in his mind, without reading or listening to the whole conversation. Based on the existing dialogue dataset, we build a new dataset, which has more than one hundred thousand dialogue-description pairs. As a step forward, we demonstrate that one can get more accurate and descriptive results using a new neural attentive model that exploits the interaction between utterances from different speakers, compared with other baselines.

Zhou Zhao, Hanbing Zhan, Lingtao Meng et al.

Retweet prediction is a challenging problem in social media sites (SMS). In this paper, we study the problem of image retweet prediction in social media, which predicts the image sharing behavior that the user reposts the image tweets from their followees. Unlike previous studies, we learn user preference ranking model from their past retweeted image tweets in SMS. We first propose heterogeneous image retweet modeling network (IRM) that exploits users' past retweeted image tweets with associated contexts, their following relations in SMS and preference of their followees. We then develop a novel attentional multi-faceted ranking network learning framework with textually guided multi-modal neural networks for the proposed heterogenous IRM network to learn the joint image tweet representations and user preference representations for prediction task. The extensive experiments on a large-scale dataset from Twitter site shows that our method achieves better performance than other state-of-the-art solutions to the problem.

Lannan Luo, Qiang Zeng, Chen Cao et al.

Android Framework is a layer of software that exists in every Android system managing resources of all Android apps. A vulnerability in Android Framework can lead to severe hacks, such as destroying user data and leaking private information. With tens of millions of Android devices unpatched due to Android fragmentation, vulnerabilities in Android Framework certainly attract attackers to exploit them. So far, enormous manual effort is needed to craft such exploits. To our knowledge, no research has been done on automatic generation of exploits that take advantage of Android Framework vulnerabilities. We make a first step towards this goal by applying symbolic execution of Android Framework to finding bugs and generating exploits. Several challenges have been raised by the task. (1) The information of an app flows to Android Framework in multiple intricate steps, making it difficult to identify symbolic inputs. (2) Android Framework has a complex initialization phase, which exacerbates the state space explosion problem. (3) A straightforward design that builds the symbolic executor as a layer inside the Android system will not work well: not only does the implementation have to ensure the compatibility with the Android system, but it needs to be maintained whenever Android gets updated. We present novel ideas and techniques to resolve the challenges, and have built the first system for symbolic execution of Android Framework. It fundamentally changes the state of the art in exploit generation on the Android system, and has been applied to constructing new techniques for finding vulnerabilities.

Yunde Jia, Bin Xu, Jiajun Shen et al.

This paper presents a telepresence interaction framework based on touchscreen and telepresence-robot technologies. The core of the framework is a new user interface, Touchable live video Image based User Interface, called TIUI. The TIUI allows a remote operator to not just drive the telepresence robot but operate and interact with real objects by touching their live video images on a pad with finger touch gestures. We implemented a telepresence interaction system which is composed of a telepresence robot and tele-interactive objects located in a local space, the TIUI of a pad located in a remote space, and the wireless networks connecting the two spaces. Our system can be a perfect embodiment of a remote operator to do most of daily living tasks, such as opening a door, drawing a curtain, pushing a wheelchair, and other like tasks. The evaluation and demonstration results show the effectiveness and promising applications of our system.

Min Yang, Yunde Jia

Robust online multi-person tracking requires the correct associations of online detection responses with existing trajectories. We address this problem by developing a novel appearance modeling approach to provide accurate appearance affinities to guide data association. In contrast to most existing algorithms that only consider the spatial structure of human appearances, we exploit the temporal dynamic characteristics within temporal appearance sequences to discriminate different persons. The temporal dynamic makes a sufficient complement to the spatial structure of varying appearances in the feature space, which significantly improves the affinity measurement between trajectories and detections. We propose a feature selection algorithm to describe the appearance variations with mid-level semantic features, and demonstrate its usefulness in terms of temporal dynamic appearance modeling. Moreover, the appearance model is learned incrementally by alternatively evaluating newly-observed appearances and adjusting the model parameters to be suitable for online tracking. Reliable tracking of multiple persons in complex scenes is achieved by incorporating the learned model into an online tracking-by-detection framework. Our experiments on the challenging benchmark MOTChallenge 2015 demonstrate that our method outperforms the state-of-the-art multi-person tracking algorithms.