Generative models are subject to overfitting and thus may potentially leak sensitive information from the training data. In this work. we investigate the privacy risks that can potentially arise from the use of generative adversarial networks (GANs) for creating tabular synthetic datasets. For the purpose, we analyse the effects of re-identification attacks on synthetic data, i.e., attacks which aim at selecting samples that are predicted to correspond to memorised training samples based on their proximity to the nearest synthetic records. We thus consider multiple settings where different attackers might have different access levels or knowledge of the generative model and predictive, and assess which information is potentially most useful for launching more successful re-identification attacks. In doing so we also consider the situation for which re-identification attacks are formulated as reconstruction attacks, i.e., the situation where an attacker uses evolutionary multi-objective optimisation for perturbing synthetic samples closer to the training space. The results indicate that attackers can indeed pose major privacy risks by selecting synthetic samples that are likely representative of memorised training samples. In addition, we notice that privacy threats considerably increase when the attacker either has knowledge or has black-box access to the generative models. We also find that reconstruction attacks through multi-objective optimisation even increase the risk of identifying confidential samples.

Hypergraphs are marked by complex topology, expressing higher-order interactions among multiple nodes with hyperedges, and better capturing the topology is essential for effective representation learning. Recent advances in generative self-supervised learning (SSL) suggest that hypergraph neural networks learned from generative self supervision have the potential to effectively encode the complex hypergraph topology. Designing a generative SSL strategy for hypergraphs, however, is not straightforward. Questions remain with regard to its generative SSL task, connection to downstream tasks, and empirical properties of learned representations. In light of the promises and challenges, we propose a novel generative SSL strategy for hypergraphs. We first formulate a generative SSL task on hypergraphs, hyperedge filling, and highlight its theoretical connection to node classification. Based on the generative SSL task, we propose a hypergraph SSL method, HypeBoy. HypeBoy learns effective general-purpose hypergraph representations, outperforming 16 baseline methods across 11 benchmark datasets.

In deep learning, test-time adaptation has gained attention as a method for model fine-tuning without the need for labeled data. A prime exemplification is the recently proposed test-time prompt tuning for large-scale vision-language models such as CLIP. Unfortunately, these prompts have been mainly developed to improve accuracy, overlooking the importance of calibration, which is a crucial aspect for quantifying prediction uncertainty. However, traditional calibration methods rely on substantial amounts of labeled data, making them impractical for test-time scenarios. To this end, this paper explores calibration during test-time prompt tuning by leveraging the inherent properties of CLIP. Through a series of observations, we find that the prompt choice significantly affects the calibration in CLIP, where the prompts leading to higher text feature dispersion result in better-calibrated predictions. Introducing the Average Text Feature Dispersion (ATFD), we establish its relationship with calibration error and present a novel method, Calibrated Test-time Prompt Tuning (C-TPT), for optimizing prompts during test-time with enhanced calibration. Through extensive experiments on different CLIP architectures and datasets, we show that C-TPT can effectively improve the calibration of test-time prompt tuning without needing labeled data. Pre-trained large-scale vision-language models, such as CLIP (Radford et al., 2021), have demonstrated the potential as foundation models by leveraging their zero-shot inference capabilities. These models are pre-trained on a massive dataset of image-text caption pairs and learn to associate the image and its corresponding text caption in a shared latent space, which allows the model to accurately classify newfound visual categories in a zero-shot manner based on carefully designed prompt templates. As hand-crafted prompts consisting of predefined vocabulary tokens (i.e., hard prompts) may not be optimal, significant attention is being directed towards prompt tuning, which treats the prompts as learnable vectors that can be optimized through gradient descent.

Enhancing generalization and uncertainty quantification in pre-trained language models (PLMs) is crucial for their effectiveness and reliability. Building on machine learning research that established the importance of robustness for improving generalization, we investigate the role of representation smoothness, achieved via Jacobian and Hessian regularization, in enhancing PLM performance. Although such regularization methods have proven effective in computer vision, their application in natural language processing (NLP), where PLM inputs are derived from a discrete domain, poses unique challenges. We introduce a novel two-phase regularization approach, JacHess, which minimizes the norms of the Jacobian and Hessian matrices within PLM intermediate representations relative to their inputs. Our evaluation using the GLUE benchmark demonstrates that JacHess significantly improves in-domain generalization and calibration in PLMs, outperforming unregularized fine-tuning and other similar regularization methods.

Legal professionals frequently encounter long legal judgments that hold critical insights for their work. While recent advances have led to automated summarization solutions for legal documents, they typically provide generic summaries, which may not meet the diverse information needs of users. To address this gap, we introduce LexAbSumm, a novel dataset designed for aspect-based summarization of legal case decisions, sourced from the European Court of Human Rights jurisdiction. We evaluate several abstractive summarization models tailored for longer documents on LexAbSumm, revealing a challenge in conditioning these models to produce aspect-specific summaries. We release LexAbSum to facilitate research in aspect-based summarization for legal domain.

We introduce methods for discovering and applying sparse feature circuits. These are causally implicated subnetworks of human-interpretable features for explaining language model behaviors. Circuits identified in prior work consist of polysemantic and difficult-to-interpret units like attention heads or neurons, rendering them unsuitable for many downstream applications. In contrast, sparse feature circuits enable detailed understanding of unanticipated mechanisms. Because they are based on fine-grained units, sparse feature circuits are useful for downstream tasks: We introduce SHIFT, where we improve the generalization of a classifier by ablating features that a human judges to be task-irrelevant. Finally, we demonstrate an entirely unsupervised and scalable interpretability pipeline by discovering thousands of sparse feature circuits for automatically discovered model behaviors.

In data-driven decision-making in marketing, healthcare, and education, it is desirable to utilize a large amount of data from existing ventures to navigate high-dimensional feature spaces and address data scarcity in new ventures. We explore knowledge transfer in dynamic decision-making by concentrating on batch stationary environments and formally defining task discrepancies through the lens of Markov decision processes (MDPs). We propose a framework of Transferred Fitted $Q$-Iteration algorithm with general function approximation, enabling the direct estimation of the optimal action-state function $Q^*$ using both target and source data. We establish the relationship between statistical performance and MDP task discrepancy under sieve approximation, shedding light on the impact of source and target sample sizes and task discrepancy on the effectiveness of knowledge transfer. We show that the final learning error of the $Q^*$ function is significantly improved from the single task rate both theoretically and empirically.

We propose a novel model-selection method for dynamic real-life networks. Our approach involves training a classifier on a large body of synthetic network data. The data is generated by simulating nine state-of-the-art random graph models for dynamic networks, with parameter range chosen to ensure exponential growth of the network size in time. We design a conceptually novel type of dynamic features that count new links received by a group of vertices in a particular time interval. The proposed features are easy to compute, analytically tractable, and interpretable. Our approach achieves a near-perfect classification of synthetic networks, exceeding the state-of-the-art by a large margin. Applying our classification method to real-world citation networks gives credibility to the claims in the literature that models with preferential attachment, fitness and aging fit real-world citation networks best, although sometimes, the predicted model does not involve vertex fitness.

Causal effect estimation aims at estimating the Average Treatment Effect as well as the Conditional Average Treatment Effect of a treatment to an outcome from the available data. This knowledge is important in many safety-critical domains, where it often needs to be extracted from observational data. In this work, we propose a new causal inference model, named C-XGBoost, for the prediction of potential outcomes. The motivation of our approach is to exploit the superiority of tree-based models for handling tabular data together with the notable property of causal inference neural network-based models to learn representations that are useful for estimating the outcome for both the treatment and non-treatment cases. The proposed model also inherits the considerable advantages of XGBoost model such as efficiently handling features with missing values requiring minimum preprocessing effort, as well as it is equipped with regularization techniques to avoid overfitting/bias. Furthermore, we propose a new loss function for efficiently training the proposed causal inference model. The experimental analysis, which is based on the performance profiles of Dolan and Mor{\'e} as well as on post-hoc and non-parametric statistical tests, provide strong evidence about the effectiveness of the proposed approach.

When estimating the direct and indirect causal effects using the influence function-based estimator of the mediation functional, it is crucial to understand what aspects of the treatment, the mediator, and the outcome mean mechanisms should be focused on. Specifically, considering them as nuisance functions and attempting to fit these nuisance functions as accurate as possible is not necessarily the best approach to take. In this work, we propose a two-stage estimation strategy for the nuisance functions that estimates the nuisance functions based on the role they play in the structure of the bias of the influence function-based estimator of the mediation functional. We provide robustness analysis of the proposed method, as well as sufficient conditions for consistency and asymptotic normality of the estimator of the parameter of interest.