WWW 2025 Papers

Retrieval-augmented generation (RAG) has demonstrated effectiveness in mitigating the hallucination problem of large language models (LLMs). However, the difficulty of aligning the retriever with the diverse LLMs' knowledge preferences inevitably poses an inevitable challenge in developing a reliable RAG system. To address this issue, we propose DPA-RAG, a universal framework designed to align diverse knowledge preferences within RAG systems. Specifically, we initially introduce a preference knowledge construction pipline and incorporate five novel query augmentation strategies to alleviate preference data scarcity. Based on preference data, DPA-RAG accomplishes both external and internal preference alignment: 1) It jointly integrate pair-wise, point-wise, and contrastive preference alignment abilities into the reranker, achieving external preference alignment among RAG components. 2) It further introduces a pre-aligned stage before vanilla Supervised Fine-tuning (SFT), enabling LLMs to implicitly capture knowledge aligned with their reasoning preferences, achieving LLMs' internal alignment. Experimental results across four knowledge-intensive QA datasets demonstrate that DPA-RAG outperforms all baselines and seamlessly integrates both black-box and open-sourced LLM readers. Further qualitative analysis and discussions also provide empirical guidance for achieving reliable RAG systems.

Item Response Theory (IRT) has been widely used in educational psychometrics to assess student ability, as well as the difficulty and discrimination of test questions. In this context, discrimination specifically refers to how effectively a question distinguishes between students of different ability levels, and it does not carry any connotation related to fairness. In recent years, IRT has been successfully used to evaluate the predictive performance of Machine Learning (ML) models, but this paper marks its first application in fairness evaluation. In this paper, we propose a novel Fair-IRT framework to evaluate a set of predictive models on a set of individuals, while simultaneously eliciting specific parameters, namely, the ability to make fair predictions (a feature of predictive models), as well as the discrimination and difficulty of individuals that affect the prediction results. Furthermore, we conduct a series of experiments to comprehensively understand the implications of these parameters for fairness evaluation. Detailed explanations for item characteristic curves (ICCs) are provided for particular individuals. We propose the flatness of ICCs to disentangle the unfairness between individuals and predictive models. The experiments demonstrate the effectiveness of this framework as a fairness evaluation tool. Two real-world case studies illustrate its potential application in evaluating fairness in both classification and regression tasks. Our paper aligns well with the Responsible Web track by proposing a Fair-IRT framework to evaluate fairness in ML models, which directly contributes to the development of a more inclusive, equitable, and trustworthy AI.

Learning-based Behavior Graph Classification (BGC) has been widely adopted in Internet infrastructure for partitioning and identifying similar behavior graphs. However, the research communities realize significant limitations when deploying existing proposals in real-world scenarios. The challenges are mainly concerned with (i) fine-grained emerging behavior graphs, and (ii) incremental model adaptations. To tackle these problems, we propose to (i) mine semantics in multi-source logs using Large Language Models (LLMs) under In-Context Learning (ICL), and (ii) bridge the gap between Out-Of-Distribution (OOD) detection and class-incremental graph learning. Based on the above core ideas, we develop the first unified framework termed as $\textbf{F}$ine-$\textbf{G}$rained and $\textbf{C}$lass-$\textbf{I}$ncremental $\textbf{B}$ehavior $\textbf{G}$raph $\textbf{C}$lassification ($\textbf{FG-CIBGC}$). It consists of two novel modules, i.e., gPartition and gAdapt, that are used for partitioning fine-grained graphs and performing unknown class detection and adaptation, respectively. To validate the efficacy of FG-CIBGC, we introduce a new benchmark, comprising a new 4,992-graph, 32-class dataset generated from 8 attack scenarios, as well as a novel Edge Intersection over Union (EIoU) metric for evaluation. Extensive experiments demonstrate FG-CIBGC's superior performance on fine-grained and class-incremental BGC tasks, as well as its ability to generate fine-grained behavior graphs that facilitate downstream tasks. The code and dataset are available at: https://anonymous.4open.science/r/FG-CIBGC-70BC/README.md.

Fairness and relevance are two important aspects of recommender systems (RSs). Typically, they are evaluated either (i) separately by individual measures of fairness and relevance, or (ii) jointly using a single measure that accounts for fairness with respect to relevance. However, approach (i) often does not provide a reliable joint estimate of the goodness of the models, as it has two different best models: one for fairness and another for relevance. Approach (ii) is also problematic because these measures tend to be ad-hoc and do not relate well to traditional relevance measures, like NDCG. Motivated by this, we present a new approach for jointly evaluating fairness and relevance in RSs: distance from pareto frontier (DPFR). Given a user-item interaction dataset, we compute their Pareto frontier for a pair of existing relevance and fairness measures, and then use the distance from the frontier as a measure of the jointly achievable fairness and relevance. Our approach is modular and intuitive as it can be computed with existing measures. Experiments with 4 RS models, 3 re-ranking strategies, and 6 datasets show that the existing metrics have inconsistent associations with our Pareto-optimal solution, making DPFR a more robust and theoretically well-founded joint measure for assessing both fairness and relevance.

The vision of Web3 is to improve user control over data and assets, but one challenge that complicates this vision is the prevalence of non-transparent, scam-prone applications and vulnerable smart contracts that put web3 users at risk. While code audits are one solution to this problem, the lack of smart contracts source code on many blockchain platforms, such as Sui, hinders the ease of auditing. A promising approach to this issue is the use of a decompiler to reverse-engineer smart contract bytecode. However, existing decompilers for Sui produce code that is difficult to understand and cannot be directly recompiled. To address this, we developed the Move AI Decompiler (MAD), a Large Language Model (LLM)-powered web application that decompiles smart contract bytecodes on Sui into logically correct, human-readable, and re-compilable source code. MAD empowers developers to understand and audit contracts easily and independently. Our evaluation shows that MAD produces logically correct code that successfully passes original unit tests and achieves a 66.7\% recompilation success rate on real-world smart contracts. Additionally, in a user study involving 12 developers, MAD significantly reduced the auditing workload compared to using traditional decompilers. Participants found MAD’s outputs comparable to the original source code, simplifying the process of smart contract logic comprehension and auditing. Despite some limitations, such as occasional hallucinations and compile errors, MAD still provides significant improvements over traditional decompilers. MAD has practical implications for blockchain smart contract transparency, auditing, and education. It empowers users to review and audit non-open-source smart contracts, fostering trust and accountability. Additionally, MAD's approach could potentially extend to other smart contract languages, like Solidity, promoting transparency across various blockchains.

The smoothing issue in graph learning leads to indistinguishable node representations, posing significant challenges for graph-related tasks. However, our experiments reveal that this problem can uncover underlying properties of node anomaly detection (NAD) that previous research has missed. We introduce Individual Smoothing Patterns (ISP) and Neighborhood Smoothing Patterns (NSP), which indicate that the representations of anomalous nodes are harder to smooth than those of normal ones. In addition, we explore the theoretical implications of these patterns, demonstrating the potential benefits of ISP and NSP for NAD tasks. Motivated by these findings, we propose SmoothGNN, a novel unsupervised NAD framework. First, we design a learning component to explicitly capture ISP for detecting node anomalies. Second, we design a spectral graph neural network to implicitly learn ISP to enhance detection. Third, we design an effective coefficient based on our findings that NSP can serve as coefficients for node representations, aiding in the identification of anomalous nodes. Furthermore, we devise a novel anomaly measure to calculate loss functions and anomalous scores for nodes, reflecting the properties of NAD using ISP and NSP. Extensive experiments on 9 real datasets show that SmoothGNN outperforms the best rival by an average of 14.66% in AUC and 7.28% in Average Precision, with 75x running time speedup, validating the effectiveness and efficiency of our framework.

Dynamic Dependence Network (DDN) inference is crucial for understanding evolving relationships in multimodal time series web data, with broad applications in fields like medical and financial network analysis. The inherent dynamic nature, temporal continuity, and heterogeneous data sources in multimodal time series data pose three fundamental challenges: computational efficiency, prediction stability and robustness, and modality quality disparity. Previous methods, generally lacking utilization of multiple modalities, either struggle with computational efficiency due to the time-intensive manual hyperparameter tuning, or compromise prediction stability and robustness by neglecting temporal coherence. To address these challenges, we propose a Normalized mutual information-driven Tuning-free Dynamic Dependence Network inference method for multimodal data, namely NoTeNet. NoTeNet provides a promising paradigm that can integrate two different data modalities to enhance prediction accuracy. It uses normalized mutual information transforms noisy auxiliary data into relationship matrices and employs a kernel function for smooth temporal estimation. Additionally, NoTeNet significantly reduces the need for manual hyperparameter adjustments, offering a tuning-free approach with theoretical guarantees. On various synthetic datasets and real-world data, NoTeNet demonstrates superior prediction accuracy and efficiency without the need for hyperparameter tuning, making it potential for a wide range of web data applications.

We study the machine unlearning problem which aims to remove specific training data from a pre-trained machine learning model to allow users to exercise their ‘right to be forgotten’ to protect user privacy. Conventional machine unlearning methods would degrade the model performance after the unlearning procedure. To mitigate the issue, they typically rely on the access to the remaining training data to fine-tune the unlearned model to mitigate the influence of unlearning. However, accessing the remaining training data may not always be practical for different reasons (e.g., data expiration policies, storage limitations, or additional privacy constraints). Machine unlearning without access to the remaining training data poses significant challenges to retaining model performance. In this paper, we study how to unlearn specific training data from a pre-trained model without accessing the remaining training data and protect model performance without dramatically changing the model's parameters. We propose a practical method called Targeted Label Noise Injection. Intuitively, our method assigns incorrect yet controllable labels to the examples that need to be forgotten and fine-tunes the pre-trained model to learn these new labels. This strategy effectively moves the to-be-forgotten examples across the decision boundary with a small impact on the model's overall performance. We theoretically prove the effectiveness of the proposed method and empirically show that it achieves state-of-the-art unlearning performance across various datasets.