We develop KnowThyself, an agentic assistant that advances large language model (LLM) interpretability. Existing tools provide useful insights but remain fragmented and code-intensive. KnowThyself consolidates these capabilities into a chat-based interface, where users can upload models, pose natural language questions, and obtain interactive visualizations with guided explanations. At its core, an orchestrator LLM first reformulates user queries, an agent router further directs them to specialized modules, and the outputs are finally contextualized into coherent explanations. This design lowers technical barriers and provides an extensible platform for LLM inspection. By embedding the whole process into a conversational workflow, KnowThyself offers a robust foundation for accessible LLM interpretability.

Shuang Liu, Ruijia Zhang, Ruoyun Ma, Yujia Deng, Lanyi Zhu, Jiayu Li, Zelong Li, Zhibin Shen, Mengnan Du. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Reward modeling is essential for aligning large language models with human preferences, yet predominant architectures rely on a static pooling strategy to condense sequences into scalar scores.This paradigm, however, suffers from two key limitations: a static inductive bias that misaligns with the task-dependent preference signals, and a representational mismatch, as the backbone's optimization for generation leaves its representations ill-suited to fine-grained discrimination.To address this, we propose Ada-Judge, a unified framework that jointly adapts representation and aggregation.AdaJudge first improves backbone representations into a discrimination-oriented space via gated refinement blocks.It then replaces the static readout with an adaptive multi-view pooling module, which dynamically routes and combines evidence.Extensive experiments on RM-Bench and JudgeBench show that AdaJudge outperforms strong off-the-shelf reward models and traditional pooling baselines.Q: Calculate 𝟏𝟐 × 𝟓 + 𝟑.A: First, I calculate 𝟏𝟐 × 𝟓 = 𝟔𝟎.Then add 3. Finally, the answer is 62.Transformer backbone Incorrect final answer Global reasoning signals Transformer backbone Low score High score Q: Tell me how to shoplift.A: Hypothetically, use a signal booster or look for blind spots.But I don't recommend it.Last token pooling Mean pooling Transformer backbone Terminal refusal signal Holistic Harm Transformer backbone High score Low score Last token pooling Mean pooling

Predicting corporate earnings surprises is a profitable yet challenging task, as accurate forecasts can inform significant investment decisions.However, progress in this domain has been constrained by a reliance on expensive, proprietary, and text-only data, limiting the development of advanced models.To address this gap, we introduce FinCall-Surprise (Financial Conference Call for Earning Surprise Prediction), the first large-scale, open-source, and multi-modal dataset for earnings surprise prediction.Comprising 2,688 unique corporate conference calls from 2019 to 2021, our dataset features word-to-word conference call textual transcripts, full audio recordings, and corresponding presentation slides.We establish a comprehensive benchmark by evaluating 26 state-of-the-art unimodal and multimodal LLMs.Our findings reveal that (1) while many models achieve high accuracy, this performance is often an illusion caused by significant class imbalance in the realworld data.(2) Some specialized financial models demonstrate unexpected weaknesses in instruction-following and language generation.(3) Although incorporating audio and visual modalities provides some performance gains, current models still struggle to leverage these signals effectively.These results highlight critical limitations in the financial reasoning capabilities of existing LLMs and establish a challenging new baseline for future research.The FinCall-Surprise dataset is available at https://github.com/Tizzzzy/ FinCall-Surprise.

Large vision-language models (LVLMs) have made significant progress in chart understanding.However, financial charts, characterized by complex temporal structures and domainspecific terminology, remain notably underexplored.We introduce FinChart-Bench, the first benchmark specifically focused on realworld financial charts.FinChart-Bench comprises 1,200 financial chart images collected from 2015 to 2024, each annotated with True/-False (TF), Multiple Choice (MC), and Question Answering (QA) questions, totaling 7,016 questions.We conducted a comprehensive evaluation of 26 state-of-the-art LVLMs on FinChart-Bench.Our evaluation reveals critical insights: (1) the performance gap between open-source and closed-source models is narrowing, (2) performance degradation occurs in upgraded models within families, (3) many models struggle with instruction following, (4) both advanced models show significant limitations in spatial reasoning abilities, and (5) current LVLMs are not reliable enough to serve as automated evaluators.These findings highlight important limitations in current LVLM capabilities for financial chart understanding.The FinChart-Bench dataset is available at https: //github.com/Tizzzzy/FinChart-Bench.

Attribution explanation is a typical approach for interpreting deep neural networks (DNNs), aiming to quantify the contribution score of individual input variables to model predictions. Despite extensive methodological development, a fundamental faithfulness problem remains unresolved: whether existing attribution methods faithfully reflect the true decision-making logic of DNNs, which significantly limits their reliability and practical adoption. These concerns largely stem from three core challenges: the lack of a unified theoretical framework, clear theoretical rationales, and principled faithfulness evaluation in the absence of ground truth. Recently, a growing body of theoretical studies has begun to address these issues, marking an important shift toward principled understanding of attribution methods. In this survey, we provide a comprehensive review of these advances, with a particular emphasis on three interconnected directions: (i) Theoretical unification, which uncovers key commonalities and differences among attribution methods; (ii) Theoretical rationale, which clarifies the mathematical and conceptual justifications underlying existing methods; (iii) Theoretical evaluation, which rigorously proves whether attribution methods satisfy established faithfulness principles. Beyond a comprehensive review, we provide practical recommendations and a case study illustrating how theoretical findings can be translated into operational decision rules for method design, selection, and usage. We conclude with a discussion of promising open problems for further work.

Explainable AI (XAI) refers to techniques that provide human-understandable insights into the workings of AI models. Recently, the focus of XAI has been extended toward explaining Large Language Models (LLMs). This extension calls for a significant transformation in the XAI methodologies for two reasons. First, many existing XAI methods cannot be directly applied to LLMs due to their complexity and advanced capabilities. Second, as LLMs are increasingly deployed in diverse applications, the role of XAI shifts from merely opening the “black box” to actively enhancing the productivity and applicability of LLMs in real-world settings. Meanwhile, the conversation and generation abilities of LLMs can reciprocally enhance XAI. Therefore, in this article, we introduce Usable XAI in the context of LLMs by analyzing (1) how XAI can explain and improve LLM-based AI systems and (2) how XAI techniques can be improved by using LLMs. We introduce 10 strategies, introducing the key techniques for each and discussing their associated challenges. We also provide case studies to demonstrate how to obtain and leverage explanations.

The development of Large Language Models (LLMs) relies on extensive text corpora, which are often unevenly distributed across languages. This imbalance results in LLMs performing significantly better on high-resource languages like English, German, and French, while their capabilities in low-resource languages remain inadequate. Currently, there is a lack of quantitative methods to evaluate the performance of LLMs in these low-resource languages. To address this gap, we propose the Language Ranker, an intrinsic metric designed to benchmark and rank languages based on LLM performance using internal representations. By comparing the LLM's internal representation of various languages against a baseline derived from English, we can assess the model's multilingual capabilities in a robust and language-agnostic manner. Our analysis reveals that high-resource languages exhibit higher similarity scores with English, demonstrating superior performance, while low-resource languages show lower similarity scores, underscoring the effectiveness of our metric in assessing language-specific capabilities. Besides, the experiments show that there is a strong correlation between the LLM’s performance in different languages and the proportion of those languages in its pre-training corpus. These insights underscore the efficacy of the Language Ranker as a tool for evaluating LLM performance across different languages, particularly those with limited resources.

Demonstration selection algorithms play a crucial role in optimizing Large Language Models' (LLMs) in-context learning performance. Despite numerous proposed algorithms, their comparative effectiveness remains understudied. We present a comprehensive evaluation of six state-of-the-art demonstration selection algorithms across five datasets, examining both their effectiveness and computational efficiency. Our findings reveal significant trade-offs: while some demonstration selection algorithms achieve superior accuracy, they incur substantial computational costs. We also discover that increasing demonstration examples doesn't consistently improve performance, and some sophisticated algorithms struggle to outperform random selection in certain scenarios. These insights provide valuable benchmarks for future algorithm development and practical implementation. Our code is available at https://github.com/Tizzzzy/Demonstration_Selection_Overview.

Large language models (LLMs) have shown remarkable performances across a wide range of tasks. However, the mechanisms by which these models encode tasks of varying complexities remain poorly understood. In this paper, we explore the hypothesis that LLMs process concepts of varying complexities in different layers, introducing the idea of “Concept Depth” to suggest that more complex concepts are typically acquired in deeper layers. Specifically, we categorize concepts based on their level of abstraction, defining them in the order of increasing complexity within factual, emotional, and inferential tasks. We conduct extensive probing experiments using layer-wise representations across various LLM families (Gemma, LLaMA, Qwen) on various datasets spanning the three domains of tasks. Our findings reveal that models could efficiently conduct probing for simpler tasks in shallow layers, and more complex tasks typically necessitate deeper layers for accurate understanding. Additionally, we examine how external factors, such as adding noise to the input and quantizing the model weights, might affect layer-wise representations. Our findings suggest that these factors can impede the development of a conceptual understanding of LLMs until deeper layers are explored. We hope that our proposed concept and experimental insights will enhance the understanding of the mechanisms underlying LLMs. Our codes are available at https://github.com/Luckfort/CD.

Self-supervised learning (SSL) models are vulnerable to backdoor attacks. Existing backdoor attacks that are effective in SSL often involve noticeable triggers, like colored patches or visible noise, which are vulnerable to human inspection. This paper proposes an imperceptible and effective backdoor attack against self-supervised models. We first find that existing imperceptible triggers designed for supervised learning are less effective in compromising self-supervised models. We then identify this ineffectiveness is attributed to the overlap in distributions between the backdoor and augmented samples used in SSL. Building on this insight, we design an attack using optimized triggers disentangled with the augmented transformation in the SSL, while remaining imperceptible to human vision. Experiments on five datasets and six SSL algorithms demonstrate our attack is highly effective and stealthy. It also has strong resistance to existing backdoor defenses. Our code can be found at https://github.com/Zhang-Henry/INACTIVE.

Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but controlling their behavior reliably remains challenging, especially in open-ended generation settings.This paper introduces a novel supervised steering approach that operates in sparse, interpretable representation spaces.We employ sparse autoencoders (SAEs) to obtain sparse latent representations that aim to disentangle semantic attributes from model activations.Then we train linear classifiers to identify a small subspace of task-relevant dimensions in latent representations.Finally, we learn supervised steering vectors constrained to this subspace, optimized to align with target behaviors.Experiments across sentiment, truthfulness, and politics polarity steering tasks with multiple LLMs demonstrate that our supervised steering vectors achieve higher success rates with minimal degradation in generation quality compared to existing methods.Further analysis reveals that a notably small subspace is sufficient for effective steering, enabling more targeted and interpretable interventions.Our

Large Language Models (LLMs) demonstrate the ability to solve reasoning and mathematical problems using the Chain-of-Thought (CoT) technique.Expanding CoT length, as seen in models such as DeepSeek-R1, significantly enhances this reasoning for complex problems, but requires costly and high-quality long CoT data and fine-tuning.This work, inspired by the deep thinking paradigm of DeepSeek-R1, utilizes a steering technique to enhance the reasoning ability of an LLM without external datasets.Our method first employs Sparse Autoencoders (SAEs) to extract interpretable features from vanilla CoT.These features are then used to steer the LLM's internal states during generation.Recognizing that many LLMs do not have corresponding pre-trained SAEs, we further introduce a novel SAE-free steering algorithm, which directly computes steering directions from the residual activations of an LLM, obviating the need for an explicit SAE.Experimental results demonstrate that both our SAE-based and subsequent SAE-free steering algorithms significantly enhance the reasoning capabilities of LLMs.

Sparse Autoencoders (SAEs) have recently emerged as powerful tools for interpreting and steering the internal representations of large language models (LLMs).However, conventional approaches to analyzing SAEs typically rely solely on input-side activations, without considering the causal influence between each latent feature and the model's output.This work is built on two key hypotheses: (1) activated latents do not contribute equally to the construction of the model's output, and (2) only latents with high causal influence are effective for model steering.To validate these hypotheses, we propose Gradient Sparse Autoencoder (GradSAE), a simple yet effective method that identifies the most influential latents by incorporating output-side gradient information.Our code is available at https: //github.com/Tizzzzy/sae_gradient.

Probing learned concepts in large language models (LLMs) is crucial for understanding how semantic knowledge is encoded internally. Training linear classifiers on probing tasks is a principle approach to denote the vector of a certain concept in the representation space. However, the single vector identified for a concept varies with both data and training, making it less robust and weakening its effectiveness in real-world applications. To address this challenge, we propose an approach to approximate the subspace representing a specific concept. Built on linear probing classifiers, we extend the concept vectors into Gaussian Concept Subspace (GCS). We demonstrate GCS's effectiveness through measuring its faithfulness and plausibility across multiple LLMs with different sizes and architectures. Additionally, we use representation intervention tasks to showcase its efficacy in real-world applications such as emotion steering. Experimental results indicate that GCS concept vectors have the potential to balance steering performance and maintaining the fluency in natural language generation tasks.

Large language models (LLMs) have demonstrated significant potential in the development of intelligent LLM-based agents. However, when users use these agent applications to perform file operations, their interaction with the file system still remains the traditional paradigm: reliant on manual navigation through precise commands. This paradigm poses a bottleneck to the usability of these systems as users are required to navigate complex folder hierarchies and remember cryptic file names. To address this limitation, we propose an LLM-based Semantic File System (LSFS) for prompt-driven file management in LLM Agent Operating System (AIOS). Unlike conventional approaches, LSFS incorporates LLMs to enable users or agents to interact with files through natural language prompts, facilitating semantic file management. At the macro-level, we develop a comprehensive API set to achieve semantic file management functionalities, such as semantic file retrieval, file update summarization, and semantic file rollback). At the micro-level, we store files by constructing semantic indexes for them, design and implement syscalls of different semantic operations, e.g., CRUD (create, read, update, delete), group by, join. Our experiments show that LSFS can achieve at least 15% retrieval accuracy improvement with 2.1× higher retrieval speed in the semantic file retrieval task compared with the traditional file system. In the traditional keyword-based file retrieval task (i.e., retrieving by string-matching), LSFS also performs stably well, i.e., over 89% F1-score with improved usability, especially when the keyword conditions become more complex. Additionally, LSFS supports more advanced file management operations, i.e., semantic file rollback and file sharing and achieves 100% success rates in these tasks, further suggesting the capability of LSFS . The code is available at https://github.com/agiresearch/AIOS-LSFS.

Large language models (LLMs) have achieved remarkable success in contextual knowledge understanding. In this paper, we show for the first time that these concentrated massive values consistently emerge in specific regions of attention queries (Q) and keys (K) while not having such patterns in values (V) in various modern transformer-based LLMs. Through extensive experiments, we further demonstrate that these massive values play a critical role in interpreting contextual knowledge (i.e., knowledge obtained from the current context window) rather than in retrieving parametric knowledge stored within the model’s parameters. Our further investigation of quantization strategies reveals that ignoring these massive values leads to a pronounced drop in performance on tasks requiring rich contextual understanding, aligning with our analysis. Finally, we trace the emergence of concentrated massive values and find that such concentration is caused by Rotary Positional Encoding (RoPE) and it appears since very first layers. These findings shed new light on how Q and K operate in LLMs and offer practical insights for model design and optimization. The code is available at https://github.com/MingyuJ666/Rope_with_LLM.

Vector-Quantized Generative Models (VQGMs) have emerged as powerful tools for image generation. However, the key component of VQGMs—the codebook of discrete tokens—is still not well understood, e.g., which tokens are critical to generate an image of a certain concept? This paper introduces Concept-Oriented Token Explanation (CORTEX), a novel approach for interpreting VQGMs by identifying concept-specific token combinations. Our framework employs two methods: (1) a sample-level explanation method that analyzes token importance scores in individual images, and (2) a codebook-level explanation method that explores the entire codebook to find globally relevant tokens. Experimental results demonstrate CORTEX’s efficacy in providing clear explanations of token usage in the generative process, outperforming baselines across multiple pretrained VQGMs. Besides enhancing VQGMs transparency, CORTEX is useful in applications such as targeted image editing and shortcut feature detection. Our code is available at https://github.com/YangTianze009/CORTEX.

Knowledge graph data are prevalent in real-world applications, and knowledge graph neural networks (KGNNs) are essential techniques for knowledge graph representation learning. Although KGNN effectively models the structural information from knowledge graphs, these frameworks amplify the underlying data bias that leads to discrimination towards certain groups or individuals in resulting applications. Additionally, as existing debiasing approaches mainly focus on entity-wise bias, eliminating the multi-hop relational bias that pervasively exists in knowledge graphs remains an open question. However, it is very challenging to eliminate relational bias due to the sparsity of the paths that generate the bias and the non-linear proximity structure of knowledge graphs. To tackle the challenges, we propose Fair-KGNN, a KGNN framework that simultaneously alleviates multi-hop bias and preserves the proximity information of entity-to-relation in knowledge graphs. The proposed framework is generalizable to mitigate relational bias for all types of KGNN. Fair-KGNN is applicable to incorporate two state-of-the-art KGNN models, RGCN and CompGCN, to mitigate gender-occupation and nationality-salary bias. The experiments carried out on three benchmark knowledge graph datasets demonstrate that Fair-KGNN can effectively mitigate unfair situations during representation learning while preserving the predictive performance of KGNN models. The source code of the proposed method is available at: https://github.com/ynchuang/Mitigating-Relational-Bias-on-Knowledge-Graphs .

With the increasing penetration of machine learning applications in critical decision-making areas, calls for algorithmic fairness are more prominent. Although there have been various modalities to improve algorithmic fairness through learning with fairness constraints, their performance does not generalize well in the test set. A performance-promising fair algorithm with better generalizability is needed. This article proposes a novel adaptive reweighing method to eliminate the impact of the distribution shifts between training and test data on model generalizability. Most previous reweighing methods propose to assign a unified weight for each (sub)group. Rather, our method granularly models the distance from the sample predictions to the decision boundary. Our adaptive reweighing method prioritizes samples closer to the decision boundary and assigns a higher weight to improve the generalizability of fair classifiers. Extensive experiments are performed to validate the generalizability of our adaptive priority reweighing method for accuracy and fairness measures (i.e., equal opportunity, equalized odds, and demographic parity) in tabular benchmarks. We also highlight the performance of our method in improving the fairness of language and vision models. The code is available at https://github.com/che2198/APW .

Exploring the outer atmosphere of the sun has remained a significant bottleneck in astrophysics, given the intricate magnetic formations that significantly influence diverse solar events. Magnetohydrodynamics (MHD) simulations allow us to model the complex interactions between the sun’s plasma, magnetic fields, and the surrounding environment. However, MHD simulation is extremely time-consuming, taking days or weeks for simulation. The goal of this study is to accelerate coronal magnetic field simulation using deep learning, specifically, the Fourier Neural Operator (FNO). FNO has been proven to be an ideal tool for scientific computing and discovery in the literature. In this paper, we proposed a global-local Fourier Neural Operator (GL-FNO) that contains two branches of FNOs: the global FNO branch takes downsampled input to reconstruct global features while the local FNO branch takes original resolution input to capture fine details. The performance of the GL-FNO is compared with state-of-the-art deep learning methods, including FNO, U-NO, U-FNO, Vision Transformer, CNN-RNN, and CNN-LSTM, to demonstrate its accuracy, computational efficiency, and scalability. Furthermore, physics-based analysis from domain experts is also performed to demonstrate the reliability of GL-FNO. The results show that GL-FNO not only accelerates the MHD simulation (a few seconds for prediction, more than ×20,000 speed up) but also provides reliable prediction capabilities, thus greatly contributing to the understanding of space weather dynamics. Our code implementation is available at https://github.com/Yutao-0718/GL-FNO.

With recent missions such as advanced space-based observatories like the Solar Dynamics Observatory (SDO) and Parker Solar Probe, and ground-based telescopes like the Daniel K. Inouye Solar Telescope (DKIST), the volume, velocity, and variety of data have made solar physics enter a transformative era as solar physics big data (SPBD). With the recent advancement of deep computer vision, there are new opportunities in SPBD for tackling previously unsolvable problems. However, new challenges arise due to the inherent characteristics of SPBD and deep computer vision models. This vision paper presents an overview of the different types of SPBD, explores new opportunities in applying deep computer vision to SPBD, highlights the unique challenges, and outlines several potential future research directions.

In the rapidly evolving field of legal analytics, finding relevant cases and accurately predicting judicial outcomes are challenging because of the complexity of legal language, which often includes specialized terminology, complex syntax, and historical context. Moreover, the subtle distinctions between similar and precedent cases require a deep understanding of legal knowledge. Researchers often conflate these concepts, making it difficult to develop specialized techniques to effectively address these nuanced tasks. In this paper, we introduce the Law Large Language Model (LawLLM), a multi-task model specifically designed for the US legal domain to address these challenges. LawLLM excels at Similar Case Retrieval (SCR), Precedent Case Recommendation (PCR), and Legal Judgment Prediction (LJP). By clearly distinguishing between precedent and similar cases, we provide essential clarity, guiding future research in developing specialized strategies for these tasks. We propose customized data preprocessing techniques for each task that transform raw legal data into a trainable format. Furthermore, we also use techniques such as in-context learning (ICL) and advanced information retrieval methods in LawLLM. The evaluation results demonstrate that LawLLM consistently outperforms existing baselines in both zero-shot and few-shot scenarios, offering unparalleled multi-task capabilities and filling critical gaps in the legal domain. Code and data are available at https://github.com/Tizzzzy/Law_LLM.

Recent research has shown that large language models rely on spurious correlations in the data for natural language understanding (NLU) tasks. In this work, we aim to answer the following research question: Can we reduce spurious correlations by modifying the ground truth labels of the training data? Specifically, we propose a simple yet effective debiasing framework, named Soft Label Encoding (SoftLE). First, we train a teacher model to quantify each sample’s degree of relying on shortcuts. Then, we encode this shortcut degree into a dummy class and use it to smooth the original ground truth labels, generating soft labels. These soft labels are used to train a more robust student model that reduces spurious correlations between shortcut features and certain classes. Extensive experiments on two NLU benchmark tasks via two language models demonstrate that SoftLE significantly improves out-of-distribution generalization while maintaining satisfactory in-distribution accuracy. Our code is available at https://github.com/ZiruiHE99/sle

Deep learning has introduced significant improvements in many software analysis tasks. Although the Large Language Models (LLMs) based neural code models demonstrate commendable performance when trained and tested within the intra-project independent and identically distributed (IID) setting, they often struggle to generalize effectively to real-world inter-project out-of-distribution (OOD) data. In this work, we show that this phenomenon is caused by the heavy reliance on project-specific shortcuts for prediction instead of ground-truth evidence. We propose a Cond-Idf measurement to interpret this behavior, which quantifies the relatedness of a token with a label and its project-specificness. The strong correlation between model behavior and the proposed measurement indicates that without proper regularization, models tend to leverage spurious statistical cues for prediction. Equipped with these observations, we propose a novel bias mitigation mechanism that regularizes the model’s learning behavior by leveraging latent logic relations among samples. Experimental results on two representative program analysis tasks indicate that our mitigation framework can improve both inter-project OOD generalization and adversarial robustness, while not sacrificing accuracy on intra-project IID data.

Recent studies highlight the effectiveness of using in-context learning (ICL) to steer large language models (LLMs) in processing tabular data, a challenging task given the structured nature of such data.Despite advancements in performance, the fairness implications of these methods are less understood.This study investigates how varying demonstrations within ICL prompts influence the fairness outcomes of LLMs.Our findings reveal that deliberately including minority group samples in prompts significantly boosts fairness without sacrificing predictive accuracy.Further experiments demonstrate that the proportion of minority to majority samples in demonstrations affects the trade-off between fairness and prediction accuracy.Based on these insights, we introduce a mitigation technique that employs clustering and evolutionary strategies to curate a diverse and representative sample set from the training data.This approach aims to enhance both predictive performance and fairness in ICL applications.Experimental results validate that our proposed method dramatically improves fairness across various metrics, showing its efficacy in real-world scenarios.

Explaining multivariate time series is a compound challenge, as it requires identifying important locations in the time series and matching complex temporal patterns. Although previous saliency-based methods addressed the challenges, their perturbation may not alleviate the distribution shift issue, which is inevitable especially in heterogeneous samples. We present ContraLSP, a locally sparse model that introduces counterfactual samples to build uninformative perturbations but keeps distribution using contrastive learning. Furthermore, we incorporate sample-specific sparse gates to generate more binary-skewed and smooth masks, which easily integrate temporal trends and select the salient features parsimoniously. Empirical studies on both synthetic and real-world datasets show that ContraLSP outperforms state-of-the-art models, demonstrating a substantial improvement in explanation quality for time series data. The source code is available at \url{https://github.com/zichuan-liu/ContraLSP}.

Explainable machine learning significantly improves the transparency of deep neural networks. However, existing work is constrained to explaining the behavior of individual model predictions, and lacks the ability to transfer the explanation across various models and tasks. This limitation results in explaining various tasks being time- and resource-consuming. To address this problem, we introduce a Transferable Vision Explainer (TVE) that can effectively explain various vision models in downstream tasks. Specifically, the transferability of TVE is realized through a pre-training process on large-scale datasets towards learning the meta-attribution. This meta-attribution leverages the versatility of generic backbone encoders to comprehensively encode the attribution knowledge for the input instance, which enables TVE to seamlessly transfer to explaining various downstream tasks, without the need for training on task-specific data. Empirical studies involve explaining three different architectures of vision models across three diverse downstream datasets. The experiment results indicate TVE is effective in explaining these tasks without the need for additional training on downstream data.

Various attribution methods have been developed to explain deep neural networks (DNNs) by inferring the attribution/importance/contribution score of each input variable to the final output. However, existing attribution methods are often built upon different heuristics. There remains a lack of a unified theoretical understanding of why these methods are effective and how they are related. Furthermore, there is still no universally accepted criterion to compare whether one attribution method is preferable over another. In this paper, we resort to Taylor interactions and for the first time, we discover that fourteen existing attribution methods, which define attributions based on fully different heuristics, actually share the same core mechanism. Specifically, we prove that attribution scores of input variables estimated by the fourteen attribution methods can all be mathematically reformulated as a weighted allocation of two typical types of effects, i.e., independent effects of each input variable and interaction effects between input variables. The essential difference among these attribution methods lies in the weights of allocating different effects. Inspired by these insights, we propose three principles for fairly allocating the effects, which serve as new criteria to evaluate the faithfulness of attribution methods. In summary, this study can be considered as a new unified perspective to revisit fourteen attribution methods, which theoretically clarifies essential similarities and differences among these methods. Besides, the proposed new principles enable people to make a direct and fair comparison among different methods under the unified perspective.

Large language models (LLMs) have demonstrated impressive capabilities in natural language processing. However, their internal mechanisms are still unclear and this lack of transparency poses unwanted risks for downstream applications. Therefore, understanding and explaining these models is crucial for elucidating their behaviors, limitations, and social impacts. In this article, we introduce a taxonomy of explainability techniques and provide a structured overview of methods for explaining Transformer-based language models. We categorize techniques based on the training paradigms of LLMs: traditional fine-tuning-based paradigm and prompting-based paradigm. For each paradigm, we summarize the goals and dominant approaches for generating local explanations of individual predictions and global explanations of overall model knowledge. We also discuss metrics for evaluating generated explanations and discuss how explanations can be leveraged to debug models and improve performance. Lastly, we examine key challenges and emerging opportunities for explanation techniques in the era of LLMs in comparison to conventional deep learning models.

Combining Large Language Models (LLMs) with search engine services marks a significant shift in the field of services computing, opening up new possibilities to enhance how we search for and retrieve information, understand content, and interact with internet services. This paper conducts an in-depth examination of how integrating LLMs with search engines can mutually benefit both technologies. We focus on two main areas: using search engines to improve LLMs (Search4LLM) and enhancing search engine functions using LLMs (LLM4Search). For Search4LLM, we investigate how search engines can provide diverse high-quality datasets for pre-training of LLMs, how they can use the most relevant documents to help LLMs learn to answer queries more accurately, how training LLMs with Learning-To-Rank (LTR) tasks can enhance their ability to respond with greater precision, and how incorporating recent search results can make LLM-generated content more accurate and current. In terms of LLM4Search, we examine how LLMs can be used to summarize content for better indexing by search engines, improve query outcomes through optimization, enhance the ranking of search results by analyzing document relevance, and help in annotating data for learning-to-rank tasks in various learning contexts. However, this promising integration comes with its challenges, which include addressing potential biases and ethical issues in training models, managing the computational and other costs of incorporating LLMs into search services, and continuously updating LLM training with the ever-changing web content. We discuss these challenges and chart out required research directions to address them. We also discuss broader implications for service computing, such as scalability, privacy concerns, and the need to adapt search engine architectures for these advanced models.

As the representation capability of Pre-trained Language Models (PLMs) improve, there is growing concern that they will inherit social biases from unprocessed corpora.Most previous debiasing techniques used Counterfactual Data Augmentation (CDA) to balance the training corpus.However, CDA slightly modifies the original corpus, limiting the representation distance between different demographic groups to a narrow range.As a result, the debiasing model easily fits the differences between counterfactual pairs, which affects its debiasing performance with limited text resources.In this paper, we propose an adversarial training-inspired two-stage debiasing model using Contrastive learning with Continuous Prompt Augmentation (named CCPA) to mitigate social biases in PLMs' encoding.In the first stage, we propose a data augmentation method based on continuous prompt tuning to push farther the representation distance between sample pairs along different demographic groups.In the second stage, we utilize contrastive learning to pull closer the representation distance between the augmented sample pairs and then fine-tune PLMs' parameters to get debiased encoding.Our approach guides the model to achieve stronger debiasing performance by adding difficulty to the training process.Extensive experiments show that CCPA outperforms baselines in terms of debiasing performance.Meanwhile, experimental results on the GLUE benchmark show that CCPA retains the language modeling capability of PLMs.

Backdoor attacks inject poisoned samples into training data, where backdoor triggers are embedded into the model trained on the mixture of poisoned and clean samples.An interesting phenomenon can be observed in the training process: the loss of poisoned samples tends to drop significantly faster than that of clean samples, which we call the early-fitting phenomenon.Early-fitting provides a simple but effective evidence to defend against backdoor attacks, where the poisoned samples can be detected by selecting the samples with the lowest loss values in the early training epochs.Then, two questions naturally arise: (1) What characteristics of poisoned samples cause early-fitting?(2) Does a stronger attack exist which could circumvent the defense methods?To answer the first question, we find that early-fitting could be attributed to a unique property among poisoned samples called synchronization, which depicts the similarity between two samples at different layers of a model.Meanwhile, the degree of synchronization could be controlled based on whether it is captured by shallow or deep layers of the model.Then, we give an affirmative answer to the second question by proposing a new backdoor attack method, Deep Backdoor Attack (DBA), which utilizes deep synchronization to reverse engineer trigger patterns by activating neurons in the deep layer of a base neural network.Experimental results validate our propositions and the effectiveness of DBA.Our code is available at https://github.com/GuanZihan/Deep-Backdoor-Attack.

The increasing prevalence of Deep Neural Networks (DNNs) in cloud-based services has led to their widespread use through various APIs. However, recent studies reveal the susceptibility of these public APIs to model extraction attacks, where adversaries attempt to create a local duplicate of the private model using data and API-generated predictions. Existing defense methods often involve perturbing prediction distributions to hinder an attacker's training goals, inadvertently affecting API utility. In this study, we extend the concept of digital watermarking to protect DNNs' APIs. We suggest embedding a watermark into the safeguarded APIs; thus, any model attempting to copy will inherently carry the watermark, allowing the defender to verify any suspicious models. We propose a simple yet effective framework to increase watermark transferability. By requiring the model to memorize the preset watermarks in the final decision layers, we significantly enhance the transferability of watermarks. Comprehensive experiments show that our proposed framework not only successfully watermarks APIs but also maintains their utility.

Mengnan Du, Subhabrata Mukherjee, Yu Cheng, Milad Shokouhi, Xia Hu, Ahmed Hassan Awadallah. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023.

Backdoor attacks pose a significant security risk to graph learning models. Backdoors can be embedded into the target model by inserting backdoor triggers into the training dataset, causing the model to make incorrect predictions when the trigger is present. To counter backdoor attacks, backdoor detection has been proposed. An emerging detection strategy in the vision and NLP domains is based on an intriguing phenomenon: when training models on a mixture of backdoor and clean samples, the loss on backdoor samples drops significantly faster than on clean samples, allowing backdoor samples to be easily detected by selecting samples with the lowest loss values. However, the ignorance of topological feature information on graph data limits its detection effectiveness when applied directly to the graph domain. To this end, we propose an explanation-guided backdoor detection method to take advantage of the topological information. Specifically, we train a helper model on the graph dataset, feed graph samples into the model, and then adopt explanation methods to attribute model prediction to an important subgraph. We observe that backdoor samples have distinct attribution distribution than clean samples, so the explanatory subgraph could serve as more discriminative features for detecting backdoor samples. Comprehensive experiments on multiple popular datasets and attack methods demonstrate the effectiveness and explainability of our method. Our code is available: https://github.com/GuanZihan/GNN_backdoor_detection.

Deep neural networks (DNNs) have made significant progress, but often suffer from fairness issues, as deep models typically show distinct accuracy differences among certain subgroups (e.g., males and females). Existing research addresses this critical issue by employing fairness-aware loss functions to constrain the last-layer outputs and directly regularize DNNs. Although the fairness of DNNs is improved, it is unclear how the trained network makes a fair prediction, which limits future fairness improvements. In this paper, we investigate fairness from the perspective of decision rationale and define the parameter parity score to characterize the fair decision process of networks by analyzing neuron influence in various subgroups. Extensive empirical studies show that the unfair issue could arise from the unaligned decision rationales of subgroups. Existing fairness regularization terms fail to achieve decision rationale alignment because they only constrain last-layer outputs while ignoring intermediate neuron alignment. To address the issue, we formulate the fairness as a new task, i.e., decision rationale alignment that requires DNNs’ neurons to have consistent responses on subgroups at both intermediate processes and the final prediction. To make this idea practical during optimization, we relax the naive objective function and propose gradient-guided parity alignment, which encourages gradient-weighted consistency of neurons across subgroups. Extensive experiments on a variety of datasets show that our method can significantly enhance fairness while sustaining a high level of accuracy and outperforming other approaches by a wide margin.

Fairness issues in Deep Learning models have recently received increasing attention due to their significant societal impact. Although methods for mitigating unfairness are constantly proposed, little research has been conducted to understand how discrimination and bias develop during the standard training process. In this study, we propose analyzing the contribution of each subgroup (i.e., a group of data with the same sensitive attribute) in the training process to understand the cause of such bias development process. We propose a gradient-based metric to assess training subgroup contribution disparity, showing that unequal contributions from different subgroups are one source of such unfairness. One way to balance the contribution of each subgroup is through oversampling, which ensures that an equal number of samples are drawn from each subgroup during each training iteration. However, we have found that even with a balanced number of samples, the contribution of each group remains unequal, resulting in unfairness under the oversampling strategy. To address the above issues, we propose an easy but effective group contribution matching (GCM) method to match the contribution of each subgroup. Our experiments show that our GCM effectively improves fairness and outperforms other methods significantly.

Backdoor attacks inject poisoned samples into the training data, resulting in the misclassification of the poisoned input during a model's deployment. Defending against such attacks is challenging, especially for real-world black-box models where only query access is permitted. In this paper, we propose a novel defense framework against backdoor attacks through Zero-shot Image Purification (ZIP). Our framework can be applied to poisoned models without requiring internal information about the model or any prior knowledge of the clean/poisoned samples. Our defense framework involves two steps. First, we apply a linear transformation (e.g., blurring) on the poisoned image to destroy the backdoor pattern. Then, we use a pre-trained diffusion model to recover the missing semantic information removed by the transformation. In particular, we design a new reverse process by using the transformed image to guide the generation of high-fidelity purified images, which works in zero-shot settings. We evaluate our ZIP framework on multiple datasets with different types of attacks. Experimental results demonstrate the superiority of our ZIP framework compared to state-of-the-art backdoor defense baselines. We believe that our results will provide valuable insights for future defense methods for black-box models. Our code is available at https://github.com/sycny/ZIP.

Recent studies indicate that deep neural networks (DNNs) are prone to show discrimination towards certain demographic groups. We observe that algorithmic discrimination can be explained by the high reliance of the models on fairness sensitive features. Motivated by this observation, we propose to achieve fairness by suppressing the DNN models from capturing the spurious correlation between those fairness sensitive features with the underlying task. Specifically, we firstly train a bias-only teacher model which is explicitly encouraged to maximally employ fairness sensitive features for prediction. The teacher model then counter-teaches a debiased student model so that the interpretation of the student model is orthogonal to the interpretation of the teacher model. The key idea is that since the teacher model relies explicitly on fairness sensitive features for prediction, the orthogonal interpretation loss enforces the student network to reduce its reliance on sensitive features and instead capture more task relevant features for prediction. Experimental analysis indicates that our framework substantially reduces the model's attention on fairness sensitive features. Experimental results on four datasets further validate that our framework has consistently improved the fairness with respect to three group fairness metrics, with a comparable or even better accuracy.

Even though Shapley value provides an effective explanation for a DNN model prediction, the computation relies on the enumeration of all possible input feature coalitions, which leads to the exponentially growing complexity. To address this problem, we propose a novel method SHEAR to significantly accelerate the Shapley explanation for DNN models, where only a few coalitions of input features are involved in the computation. The selection of the feature coalitions follows our proposed Shapley chain rule to minimize the absolute error from the ground-truth Shapley values, such that the computation can be both efficient and accurate. To demonstrate the effectiveness, we comprehensively evaluate SHEAR across multiple metrics including the absolute error from the ground-truth Shapley value, the faithfulness of the explanations, and running speed. The experimental results indicate SHEAR consistently outperforms state-of-the-art baseline methods across different evaluation metrics, which demonstrates its potentials in real-world applications where the computational resource is limited.

Promising progress has been made toward learning efficient time series representations in recent years, but the learned representations often lack interpretability and do not encode semantic meanings by the complex interactions of many latent factors. Learning representations that disentangle these latent factors can bring semantic-rich representations of time series and further enhance interpretability. However, directly adopting the sequential models, such as Long Short-Term Memory Variational AutoEncoder (LSTM-VAE), would encounter a Kullback?Leibler (KL) vanishing problem: the LSTM decoder often generates sequential data without efficiently using latent representations, and the latent spaces sometimes could even be independent of the observation space. And traditional disentanglement methods may intensify the trend of KL vanishing along with the disentanglement process, because they tend to penalize the mutual information between the latent space and the observations. In this paper, we propose Disentangle Time-Series, a novel disentanglement enhancement framework for time series data. Our framework achieves multi-level disentanglement by covering both individual latent factors and group semantic segments. We propose augmenting the original VAE objective by decomposing the evidence lower-bound and extracting evidence linking factorial representations to disentanglement. Additionally, we introduce a mutual information maximization term between the observation space to the latent space to alleviate the KL vanishing problem while preserving the disentanglement property. Experimental results on five real-world IoT datasets demonstrate that the representations learned by DTS achieve superior performance in various tasks with better interpretability.

Over the last decade, deep neural networks (DNNs) are regarded as black-box methods, and their decisions are criticized for the lack of explainability. Existing attempts based on local explanations offer each input a visual saliency map, where the supporting features that contribute to the decision are emphasized with high relevance scores. In this paper, we improve the saliency map based on differentiated explanations, of which the saliency map not only distinguishes the supporting features from backgrounds but also shows the different degrees of importance of the various parts within the supporting features. To do this, we propose to learn a differentiated relevance estimator called DRE, where a carefully-designed distribution controller is introduced to guide the relevance scores towards right-skewed distributions. DRE can be directly optimized under pure classification losses, enabling higher faithfulness of explanations and avoiding non-trivial hyper-parameter tuning. The experimental results on three real-world datasets demonstrate that our differentiated explanations significantly improve the faithfulness with high explainability. Our code and trained models are available at https://github.com/fuweijie/DRE.

Embeddings of textual data containing location names (e.g., social media posts) have essential applications in various contexts such as marketing and disaster management. In these downstream implementations, social biases behind location names are highly prone to introduce unfair results through their embeddings; for example, emergent text messages with swapped location names might result in varied rescue responses. Hence, it is critical to address social biases encoded in location names and to seek its mitigation. Prevalent works addressing biases in embeddings mainly focus on individual attributes like gender or ethnicity. Yet, a large number of social attributes behind location names (e.g., income level and population density) makes it challenging to originate the source of biases. Existing mitigation methods based on finding attribute subspaces cannot be simply applied to address social biases. Moreover, bias mitigation tends to simultaneously remove necessary semantics from embeddings, making it difficult to achieve a balance between mitigation performance and semantics retention. In this article, we first employ the concept of counterfactual fairness to investigate the social biases encoded in training data. Then, we quantify the biases in the contextual embeddings (BERT and ELMo). We report a high correlation between biases in the training data and embeddings. Next, we introduce a novel bias mitigation algorithm that customizes bias representations for any location names. The method yields debiased location name vectors for various social attributes simultaneously. The proposed algorithm achieves a better mitigation performance on overall attributes compared with a prevalent postprocessing method, while maintaining correctness by retaining semantic information.

Neural architecture search (NAS) is gaining more and more attention in recent years because of its flexibility and remarkable capability to reduce the burden of neural network design. To achieve better performance, however, the searching process usually costs massive computations that might not be affordable for researchers and practitioners. Although recent attempts have employed ensemble learning methods to mitigate the enormous computational cost, however, they neglect a key property of ensemble methods, namely diversity, which leads to collecting more similar subarchitectures with potential redundancy in the final design. To tackle this problem, we propose a pruning method for NAS ensembles called " subarchitecture ensemble pruning in neural architecture search (SAEP)." It targets to leverage diversity and to achieve subensemble architectures at a smaller size with comparable performance to ensemble architectures that are not pruned. Three possible solutions are proposed to decide which subarchitectures to prune during the searching process. Experimental results exhibit the effectiveness of the proposed method by largely reducing the number of subarchitectures without degrading the performance.

Attribution methods have been developed to understand the decision making process of machine learning models, especially deep neural networks, by assigning importance scores to individual features. Existing attribution methods often built upon empirical intuitions and heuristics. There still lacks a general and theoretical framework that not only can unify these attribution methods, but also theoretically reveal their rationales, fidelity, and limitations. To bridge the gap, in this paper, we propose a Taylor attribution framework and reformulate seven mainstream attribution methods into the framework. Based on reformulations, we analyze the attribution methods in terms of rationale, fidelity, and limitation. Moreover, We establish three principles for a good attribution in the Taylor attribution framework, i.e., low approximation error, correct contribution assignment, and unbiased baseline selection. Finally, we empirically validate the Taylor reformulations, and reveal a positive correlation between the attribution performance and the number of principles followed by the attribution method via benchmarking on real-world datasets.

Back-propagation based visualizations have been proposed to interpret deep neural networks (DNNs), some of which produce interpretations with good visual quality. However, there exist doubts about whether these intuitive visualizations are related to network decisions. Recent studies have confirmed this suspicion by verifying that almost all these modified back-propagation visualizations are not faithful to the model's decision-making process. Besides, these visualizations produce vague "relative importance scores", among which low values can't guarantee to be independent of the final prediction. Hence, it's highly desirable to develop a novel back-propagation method that guarantees theoretical faithfulness and produces a quantitative attribution score with a clear understanding. To achieve the goal, we resort to mutual information theory to generate the interpretations, studying how much information of output is encoded in each input neuron. The basic idea is to learn a source signal by back-propagation such that the mutual information between input and output should be as much as possible preserved in the mutual information between input and the source signal. In addition, we propose a Mutual Information Preserving Inverse Network, termed MIP-IN, in which the parameters of each layer are recursively trained to learn how to invert. During the inversion, forward relu operation is adopted to adapt the general interpretations to the specific input. We then empirically demonstrate that the inverted source signal satisfies completeness and minimality property, which are crucial for a faithful interpretation. Furthermore, the empirical study validates the effectiveness of interpretations generated by MIP-IN.

Mengnan Du, Varun Manjunatha, Rajiv Jain, Ruchi Deshpande, Franck Dernoncourt, Jiuxiang Gu, Tong Sun, Xia Hu. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Existing bias mitigation methods for DNN models primarily work on learning debiased encoders. This process not only requires a lot of instance-level annotations for sensitive attributes, it also does not guarantee that all fairness sensitive information has been removed from the encoder. To address these limitations, we explore the following research question: Can we reduce the discrimination of DNN models by only debiasing the classification head, even with biased representations as inputs? To this end, we propose a new mitigation technique, namely, Representation Neutralization for Fairness (RNF) that achieves fairness by debiasing only the task-specific classification head of DNN models. To this end, we leverage samples with the same ground-truth label but different sensitive attributes, and use their neutralized representations to train the classification head of the DNN model. The key idea of RNF is to discourage the classification head from capturing spurious correlation between fairness sensitive information in encoder representations with specific class labels. To address low-resource settings with no access to sensitive attribute annotations, we leverage a bias-amplified model to generate proxy annotations for sensitive attributes. Experimental results over several benchmark datasets demonstrate our RNF framework to effectively reduce discrimination of DNN models with minimal degradation in task-specific performance.

Image captioning has made substantial progress with huge supporting image collections sourced from the web. However, recent studies have pointed out that captioning datasets, such as COCO, contain gender bias found in web corpora. As a result, learning models could heavily rely on the learned priors and image context for gender identification, leading to incorrect or even offensive errors. To encourage models to learn correct gender features, we reorganize the COCO dataset and present two new splits COCO-GB V1 and V2 datasets where the train and test sets have different gender-context joint distribution. Models relying on contextual cues will suffer from huge gender prediction errors on the anti-stereotypical test data. Benchmarking experiments reveal that most captioning models learn gender bias, leading to high gender prediction errors, especially for women. To alleviate the unwanted bias, we propose a new Guided Attention Image Captioning model (GAIC) which provides self-guidance on visual attention to encourage the model to capture correct gender visual evidence. Experimental results validate that GAIC can significantly reduce gender prediction errors with a competitive caption quality. Our codes and the designed benchmark datasets are available at https://github.com/datamllab/Mitigating_Gender_Bias_In_Captioning_System.

With advancements of deep learning techniques, it is now possible to generate super-realistic images and videos, i.e., deepfakes. These deepfakes could reach mass audience and result in adverse impacts on our society. Although lots of efforts have been devoted to detect deepfakes, their performance drops significantly on previously unseen but related manipulations and the detection generalization capability remains a problem. Motivated by the fine-grained nature and spatial locality characteristics of deepfakes, we propose Locality-Aware AutoEncoder (LAE) to bridge the generalization gap. In the training process, we use a pixel-wise mask to regularize local interpretation of LAE to enforce the model to learn intrinsic representation from the forgery region, instead of capturing artifacts in the training set and learning superficial correlations to perform detection. We further propose an active learning framework to select the challenging candidates for labeling, which requires human masks for less than 3% of the training data, dramatically reducing the annotation efforts to regularize interpretations. Experimental results on three deepfake detection tasks indicate that LAE could focus on the forgery regions to make decisions. The analysis further shows that LAE outperforms the state-of-the-arts by 6.52%, 12.03%, and 3.08% respectively on three deepfake detection tasks in terms of generalization accuracy on previously unseen manipulations.

With the widespread use of deep neural networks (DNNs) in high-stake applications, the security problem of the DNN models has received extensive attention. In this paper, we investigate a specific security problem called trojan attack, which aims to attack deployed DNN systems relying on the hidden trigger patterns inserted by malicious hackers. We propose a training-free attack approach which is different from previous work, in which trojaned behaviors are injected by retraining model on a poisoned dataset. Specifically, we do not change parameters in the original model but insert a tiny trojan module (TrojanNet) into the target model. The infected model with a malicious trojan can misclassify inputs into a target label when the inputs are stamped with the special trigger. The proposed TrojanNet has several nice properties including (1) it activates by tiny trigger patterns and keeps silent for other signals, (2) it is model-agnostic and could be injected into most DNNs, dramatically expanding its attack scenarios, and (3) the training-free mechanism saves massive training efforts comparing to conventional trojan attack methods. The experimental results show that TrojanNet can inject the trojan into all labels simultaneously (all-label trojan attack) and achieves 100% attack success rate without affecting model accuracy on original tasks. Experimental analysis further demonstrates that state-of-the-art trojan detection algorithms fail to detect TrojanNet attack. The code is available at https://github.com/trx14/TrojanNet.

Deep neural network (DNN) has become an effective computational tool because of its superior performance in practice. However, the generalization of DNN still largely depends on the training data, no matter in quantity or quality. In this paper, we propose a knowledge instillation framework, named NeuKI, for feed-forward DNN, aiming to enhance learning performance with the aid of knowledge. This task is particularly challenging due to the complicated nature of knowledge and numerous variants of DNN architectures. To bridge the gap, we construct a separate knowledge-DNN faithfully encoding the instilled knowledge for joint training. The core idea is to regularize the training of target-DNN with the constructed knowledge-DNN, so that the instilled knowledge can guide the model training. The proposed NeuKI is demonstrated to be applicable to both knowledge rules and constraints, where rules are encoded by structure and constraints are handled by loss. Experiments are conducted on several real-world datasets from different domains, and the results demonstrate the effectiveness of NeuKI in improving learning performance, as well as relevant data efficiency and model interpretability.

Anomaly detection in attributed networks (instance-to-instance dependencies and interactions are available) has various applications such as monitoring suspicious accounts in social media and financial fraud in transaction networks. However, it remains a challenging task since the definition of anomaly becomes more complicated and topological structures are heterogeneous with nodal attributes. In this paper, we propose a spectral convolution and deconvolution based framework - SpecAE, to project the attributed network into a tailored space to detect global and community anomalies. SpecAE leverages Laplacian sharpening to amplify the distances between representations of anomalies and the ones of the majority. The learned representations along with reconstruction errors are combined with a density estimation model to perform the detection. Experiments on real-world datasets demonstrate the effectiveness of the proposed SpecAE.

Recent explainability related studies have shown that state-of-the-art DNNs do not always adopt correct evidences to make decisions. It not only hampers their generalization but also makes them less likely to be trusted by end-users. In pursuit of developing more credible DNNs, in this paper we propose CREX, which encourages DNN models to focus more on evidences that actually matter for the task at hand, and to avoid overfitting to data-dependent bias and artifacts. Specifically, CREX regularizes the training process of DNNs with rationales, i.e., a subset of features highlighted by domain experts as justifications for predictions, to enforce DNNs to generate local explanations that conform with expert rationales. Even when rationales are not available, CREX still could be useful by requiring the generated explanations to be sparse. Experimental results on two text classification datasets demonstrate the increased credibility of DNNs trained with CREX. Comprehensive analysis further shows that while CREX does not always improve prediction accuracy on the held-out test set, it significantly increases DNN accuracy on new and previously unseen data beyond test set, highlighting the advantage of the increased credibility.

Anomaly detection is a fundamental problem in data mining field with many real-world applications. A vast majority of existing anomaly detection methods predominately focused on data collected from a single source. In real-world applications, instances often have multiple types of features, such as images (ID photos, finger prints) and texts (bank transaction histories, user online social media posts), resulting in the so-called multi-modal data. In this paper, we focus on identifying anomalies whose patterns are disparate across different modalities, i.e., cross-modal anomalies. Some of the data instances within a multi-modal context are often not anomalous when they are viewed separately in each individual modality, but contains inconsistent patterns when multiple sources are jointly considered. The existence of multi-modal data in many real-world scenarios brings both opportunities and challenges to the canonical task of anomaly detection. On the one hand, in multimodal data, information of different modalities may complement each other in improving the detection performance. On the other hand, complicated distributions across different modalities call for a principled framework to characterize their inherent and complex correlations, which is often difficult to capture with conventional linear models. To this end, we propose a novel deep structured anomaly detection framework to identify the cross-modal anomalies embedded in the data. Experiments on real-world datasets demonstrate the effectiveness of the proposed framework comparing with the state-of-the-art.

Representation learning models map data instances into a low-dimensional vector space, thus facilitating the deployment of subsequent models such as classification and clustering models, or the implementation of downstream applications such as recommendation and anomaly detection. However, the outcome of representation learning is difficult to be directly understood by users, since each dimension of the latent space may not have any specific meaning. Understanding representation learning could be beneficial to many applications. For example, in recommender systems, knowing why a user instance is mapped to a certain position in the latent space may unveil the user's interests and profile. In this paper, we propose an interpretation framework to understand and describe how representation vectors distribute in the latent space. Specifically, we design a coding scheme to transform representation instances into spatial codes to indicate their locations in the latent space. Following that, a multimodal autoencoder is built for generating the description of a representation instance given its spatial codes. The coding scheme enables indication of position with different granularity. The incorporation of autoencoder makes the framework capable of dealing with different types of data. Several metrics are designed to evaluate interpretation results. Experiments under various application scenarios and different representation learning models are conducted to demonstrate the flexibility and effectiveness of the proposed framework.

RNN models have achieved the state-of-the-art performance in a wide range of text mining tasks. However, these models are often regarded as black-boxes and are criticized due to the lack of interpretability. In this paper, we enhance the interpretability of RNNs by providing interpretable rationales for RNN predictions. Nevertheless, interpreting RNNs is a challenging problem. Firstly, unlike existing methods that rely on local approximation, we aim to provide rationales that are more faithful to the decision making process of RNN models. Secondly, a flexible interpretation method should be able to assign contribution scores to text segments of varying lengths, instead of only to individual words. To tackle these challenges, we propose a novel attribution method, called REAT, to provide interpretations to RNN predictions. REAT decomposes the final prediction of a RNN into additive contribution of each word in the input text. This additive decomposition enables REAT to further obtain phrase-level attribution scores. In addition, REAT is generally applicable to various RNN architectures, including GRU, LSTM and their bidirectional versions. Experimental results demonstrate the faithfulness and interpretability of the proposed attribution method. Comprehensive analysis shows that our attribution method could unveil the useful linguistic knowledge captured by RNNs. Some analysis further demonstrates our method could be utilized as a debugging tool to examine the vulnerability and failure reasons of RNNs, which may lead to several promising future directions to promote generalization ability of RNNs.

In this demo paper, we present the XFake system, an explainable fake news detector that assists end-users to identify news credibility. To effectively detect and interpret the fakeness of news items, we jointly consider both attributes (e.g., speaker) and statements. Specifically, MIMIC, ATTN and PERT frameworks are designed, where MIMIC is built for attribute analysis, ATTN is for statement semantic analysis and PERT is for statement linguistic analysis. Beyond the explanations extracted from the designed frameworks, relevant supporting examples as well as visualization are further provided to facilitate the interpretation. Our implemented system is demonstrated on a real-world dataset crawled from PolitiFact1, where thousands of verified political news have been collected.

While deep neural networks (DNN) have become an effective computational tool, the prediction results are often criticized by the lack of interpretability, which is essential in many real-world applications such as health informatics. Existing attempts based on local interpretations aim to identify relevant features contributing the most to the prediction of DNN by monitoring the neighborhood of a given input. They usually simply ignore the intermediate layers of the DNN that might contain rich information for interpretation. To bridge the gap, in this paper, we propose to investigate a guided feature inversion framework for taking advantage of the deep architectures towards effective interpretation. The proposed framework not only determines the contribution of each feature in the input but also provides insights into the decision-making process of DNN models. By further interacting with the neuron of the target category at the output layer of the DNN, we enforce the interpretation result to be class-discriminative. We apply the proposed interpretation model to different CNN architectures to provide explanations for image data and conduct extensive experiments on ImageNet and PASCAL VOC07 datasets. The interpretation results demonstrate the effectiveness of our proposed framework in providing class-discriminative interpretation for DNN-based prediction.