Search

As large language models (LLMs) are increasingly deployed in diverse applications, including chatbot assistants and code generation, aligning their behavior with safety and ethical standards has become paramount. However, jailbreak attacks, which exploit vulnerabilities to elicit unintended or harmful outputs, threaten LLMs' safety significantly. In this paper, we introduce Layer-AdvPatcher, a novel methodology designed to defend against jailbreak attacks by utilizing an unlearning strategy to patch specific layers within LLMs through self-augmented datasets. Our insight is that certain layer(s), tend to produce affirmative tokens when faced with harmful prompts. By identifying these layers and adversarially exposing them to generate more harmful data, one can understand their inherent and diverse vulnerabilities to attacks. With these exposures, we then "unlearn" these issues, reducing the impact of affirmative tokens and hence minimizing jailbreak risks while keeping the model's responses to safe queries intact. We conduct extensive experiments on two models, four benchmark datasets, and multiple state-of-the-art jailbreak attacks to demonstrate the efficacy of our approach. Results indicate that our framework reduces the harmfulness and attack success rate of jailbreak attacks without compromising utility for benign queries compared to recent defense methods. Our code is publicly available at: https://github.com/oyy2000/LayerAdvPatcher, Comment: 14 pages, 4 figures, conference

Diverse presentation formats play a pivotal role in effectively conveying code and analytical processes during data analysis. One increasingly popular format is tutorial videos, particularly those based on Jupyter notebooks, which offer an intuitive interpretation of code and vivid explanations of analytical procedures. However, creating such videos requires a diverse skill set and significant manual effort, posing a barrier for many analysts. To bridge this gap, we introduce an innovative tool called NotePlayer, which connects notebook cells to video segments and incorporates a computational engine with language models to streamline video creation and editing. Our aim is to make the process more accessible and efficient for analysts. To inform the design of NotePlayer, we conducted a formative study and performed content analysis on a corpus of 38 Jupyter tutorial videos. This helped us identify key patterns and challenges encountered in existing tutorial videos, guiding the development of NotePlayer. Through a combination of a usage scenario and a user study, we validated the effectiveness of NotePlayer. The results show that the tool streamlines the video creation and facilitates the communication process for data analysts., Comment: 20 pages, UIST 2024

In healthcare, AI techniques are widely used for tasks like risk assessment and anomaly detection. Despite AI's potential as a valuable assistant, its role in complex medical data analysis often oversimplifies human-AI collaboration dynamics. To address this, we collaborated with a local hospital, engaging six physicians and one data scientist in a formative study. From this collaboration, we propose a framework integrating two-phase interactive visualization systems: one for Human-Led, AI-Assisted Retrospective Analysis and another for AI-Mediated, Human-Reviewed Iterative Modeling. This framework aims to enhance understanding and discussion around effective human-AI collaboration in healthcare., Comment: To appear at the IEEE VIS Conference 2024

The problem of pre-training data detection for large language models (LLMs) has received growing attention due to its implications in critical issues like copyright violation and test data contamination. Despite improved performance, existing methods (including the state-of-the-art, Min-K%) are mostly developed upon simple heuristics and lack solid, reasonable foundations. In this work, we propose a novel and theoretically motivated methodology for pre-training data detection, named Min-K%++. Specifically, we present a key insight that training samples tend to be local maxima of the modeled distribution along each input dimension through maximum likelihood training, which in turn allow us to insightfully translate the problem into identification of local maxima. Then, we design our method accordingly that works under the discrete distribution modeled by LLMs, whose core idea is to determine whether the input forms a mode or has relatively high probability under the conditional categorical distribution. Empirically, the proposed method achieves new SOTA performance across multiple settings. On the WikiMIA benchmark, Min-K%++ outperforms the runner-up by 6.2% to 10.5% in detection AUROC averaged over five models. On the more challenging MIMIR benchmark, it consistently improves upon reference-free methods while performing on par with reference-based method that requires an extra reference model., Comment: ICLR'25 Spotlight. Project page and code is available at https://zjysteven.github.io/mink-plus-plus/

Missing data can pose a challenge for machine learning (ML) modeling. To address this, current approaches are categorized into feature imputation and label prediction and are primarily focused on handling missing data to enhance ML performance. These approaches rely on the observed data to estimate the missing values and therefore encounter three main shortcomings in imputation, including the need for different imputation methods for various missing data mechanisms, heavy dependence on the assumption of data distribution, and potential introduction of bias. This study proposes a Contrastive Learning (CL) framework to model observed data with missing values, where the ML model learns the similarity between an incomplete sample and its complete counterpart and the dissimilarity between other samples. Our proposed approach demonstrates the advantages of CL without requiring any imputation. To enhance interpretability, we introduce CIVis, a visual analytics system that incorporates interpretable techniques to visualize the learning process and diagnose the model status. Users can leverage their domain knowledge through interactive sampling to identify negative and positive pairs in CL. The output of CIVis is an optimized model that takes specified features and predicts downstream tasks. We provide two usage scenarios in regression and classification tasks and conduct quantitative experiments, expert interviews, and a qualitative user study to demonstrate the effectiveness of our approach. In short, this study offers a valuable contribution to addressing the challenges associated with ML modeling in the presence of missing data by providing a practical solution that achieves high predictive accuracy and model interpretability., Comment: 18 pages, 11 figures. This paper is accepted by IEEE Transactions on Visualization and Computer Graphics (TVCG)

The provision of fire services plays a vital role in ensuring the safety of residents' lives and property. The spatial layout of fire stations is closely linked to the efficiency of fire rescue operations. Traditional approaches have primarily relied on mathematical planning models to generate appropriate layouts by summarizing relevant evaluation criteria. However, this optimization process presents significant challenges due to the extensive decision space, inherent conflicts among criteria, and decision-makers' preferences. To address these challenges, we propose FSLens, an interactive visual analytics system that enables in-depth evaluation and rational optimization of fire station layout. Our approach integrates fire records and correlation features to reveal fire occurrence patterns and influencing factors using spatiotemporal sequence forecasting. We design an interactive visualization method to explore areas within the city that are potentially under-resourced for fire service based on the fire distribution and existing fire station layout. Moreover, we develop a collaborative human-computer multi-criteria decision model that generates multiple candidate solutions for optimizing firefighting resources within these areas. We simulate and compare the impact of different solutions on the original layout through well-designed visualizations, providing decision-makers with the most satisfactory solution. We demonstrate the effectiveness of our approach through one case study with real-world datasets. The feedback from domain experts indicates that our system helps them to better identify and improve potential gaps in the current fire station layout., Comment: Accepted by IEEE VIS 2023

Simulation-based Medical Education (SBME) has been developed as a cost-effective means of enhancing the diagnostic skills of novice physicians and interns, thereby mitigating the need for resource-intensive mentor-apprentice training. However, feedback provided in most SBME is often directed towards improving the operational proficiency of learners, rather than providing summative medical diagnoses that result from experience and time. Additionally, the multimodal nature of medical data during diagnosis poses significant challenges for interns and novice physicians, including the tendency to overlook or over-rely on data from certain modalities, and difficulties in comprehending potential associations between modalities. To address these challenges, we present DiagnosisAssistant, a visual analytics system that leverages historical medical records as a proxy for multimodal modeling and visualization to enhance the learning experience of interns and novice physicians. The system employs elaborately designed visualizations to explore different modality data, offer diagnostic interpretive hints based on the constructed model, and enable comparative analyses of specific patients. Our approach is validated through two case studies and expert interviews, demonstrating its effectiveness in enhancing medical training., Comment: Accepted by IEEE VIS 2023

It is commonly believed that financial crises "lead to" lower growth of a country during the two-year recession period, which can be reflected by their post-crisis GDP growth. However, by contrasting a causal model with a standard prediction model, this paper argues that such a belief is non-causal. To make causal inferences, we design a two-stage staggered difference-in-differences model to estimate the average treatment effects. Interpreting the residuals as the contribution of each crisis to the treatment effects, we astonishingly conclude that cross-sectional crises are often limited to providing relevant causal information to policymakers., Comment: 12 pages, 3 figures, 5 regressions, 1 conclusion

Restructuring the regional industrial structure (RIS) has the potential to halt economic recession and achieve revitalization. Understanding the current status and dynamics of RIS will greatly assist in studying and evaluating the current industrial structure. Previous studies have focused on qualitative and quantitative research to rationalize RIS from a macroscopic perspective. Although recent studies have traced information at the industrial enterprise level to complement existing research from a micro perspective, the ambiguity of the underlying variables contributing to the industrial sector and its composition, the dynamic nature, and the large number of multivariant features of RIS records have obscured a deep and fine-grained understanding of RIS. To this end, we propose an interactive visualization system, RISeer, which is based on interpretable machine learning models and enhanced visualizations designed to identify the evolutionary patterns of the RIS and facilitate inter-regional inspection and comparison. Two case studies confirm the effectiveness of our approach, and feedback from experts indicates that RISeer helps them to gain a fine-grained understanding of the dynamics and evolution of the RIS., Comment: IEEE Transactions on Visualization and Computer Graphics (Proc. IEEE VIS 2022)

The $p_0$ model is an exponential random graph model for directed networks with the bi-degree sequence as the exclusively sufficient statistic. It captures the network feature of degree heterogeneity. The consistency and asymptotic normality of a differentially private estimator of the parameter in the private $p_0$ model has been established. However, the $p_0$ model only focuses on binary edges. In many realistic networks, edges could be weighted, taking a set of finite discrete values. In this paper, we further show that the moment estimators of the parameters based on the differentially private bi-degree sequence in the weighted $p_0$ model are consistent and asymptotically normal. Numerical studies demonstrate our theoretical findings., Comment: 19 pages. arXiv admin note: text overlap with arXiv:1705.01715 and arXiv:1408.1156 by other authors