Machine learning algorithms are increasingly provided as black-box cloud services or pre-trained models, without access to their training data. This motivates the problem of zero-shot out-of-distribution (OOD) detection. Concretely, we aim to detect OOD objects that do not belong to the classifier's label set but are erroneously classified as in-distribution (ID) objects. Our approach, RONIN, uses an off-the-shelf diffusion model to replace detected objects with inpainting. RONIN conditions the inpainting process with the predicted ID label, drawing the input object closer to the in-distribution domain. As a result, the reconstructed object is very close to the original in the ID cases and far in the OOD cases, allowing RONIN to effectively distinguish ID and OOD samples. Throughout extensive experiments, we demonstrate that RONIN achieves competitive results compared to previous approaches across several datasets, both in zero-shot and non-zero-shot settings.

The advent of generative AI images has completely disrupted the art world. Distinguishing AI generated images from human art is a challenging problem whose impact is growing over time. A failure to address this problem allows bad actors to defraud individuals paying a premium for human art and companies whose stated policies forbid AI imagery. It is also critical for content owners to establish copyright, and for model trainers interested in curating training data in order to avoid potential model collapse. There are several different approaches to distinguishing human art from AI images, including classifiers trained by supervised learning, research tools targeting diffusion models, and identification by professional artists using their knowledge of artistic techniques. In this paper, we seek to understand how well these approaches can perform against today's modern generative models in both benign and adversarial settings. We curate real human art across 7 styles, generate matching images from 5 generative models, and apply 8 detectors (5 automated detectors and 3 different human groups including 180 crowdworkers, 4000+ professional artists, and 13 expert artists experienced at detecting AI). Both Hive and expert artists do very well, but make mistakes in different ways (Hive is weaker against adversarial perturbations while Expert artists produce higher false positives). We believe these weaknesses will remain as models continue to evolve, and use our data to demonstrate why a combined team of human and automated detectors provides the best combination of accuracy and robustness.

Purpose: To develop and evaluate a semi-supervised learning model for intracranial hemorrhage detection and segmentation on an out-of-distribution head CT evaluation set. Materials and Methods: This retrospective study used semi-supervised learning to bootstrap performance. An initial "teacher" deep learning model was trained on 457 pixel-labeled head CT scans collected from one US institution from 2010-2017 and used to generate pseudo-labels on a separate unlabeled corpus of 25000 examinations from the RSNA and ASNR. A second "student" model was trained on this combined pixel- and pseudo-labeled dataset. Hyperparameter tuning was performed on a validation set of 93 scans. Testing for both classification (n=481 examinations) and segmentation (n=23 examinations, or 529 images) was performed on CQ500, a dataset of 481 scans performed in India, to evaluate out-of-distribution generalizability. The semi-supervised model was compared with a baseline model trained on only labeled data using area under the receiver operating characteristic curve (AUC), Dice similarity coefficient (DSC), and average precision (AP) metrics. Results: The semi-supervised model achieved statistically significantly higher examination AUC on CQ500 compared with the baseline (0.939 [0.938, 0.940] vs. 0.907 [0.906, 0.908]) (p=0.009). It also achieved a higher DSC (0.829 [0.825, 0.833] vs. 0.809 [0.803, 0.812]) (p=0.012) and Pixel AP (0.848 [0.843, 0.853]) vs. 0.828 [0.817, 0.828]) compared to the baseline. Conclusion: The addition of unlabeled data in a semi-supervised learning framework demonstrates stronger generalizability potential for intracranial hemorrhage detection and segmentation compared with a supervised baseline.

Smartphones have become indispensable in our daily lives and can do almost everything, from communication to online shopping. However, with the increased usage, cybercrime aimed at mobile devices is rocketing. Smishing attacks, in particular, have observed a significant upsurge in recent years. This problem is further exacerbated by the perpetrator creating new deceptive websites daily, with an average life cycle of under 15 hours. This renders the standard practice of keeping a database of malicious URLs ineffective. To this end, we propose a novel on-device pipeline: COPS that intelligently identifies features of fraudulent messages and URLs to alert the user in real-time. COPS is a lightweight pipeline with a detection module based on the Disentangled Variational Autoencoder of size 3.46MB for smishing and URL phishing detection, and we benchmark it on open datasets. We achieve an accuracy of 98.15% and 99.5%, respectively, for both tasks, with a false negative and false positive rate of a mere 0.037 and 0.015, outperforming previous works with the added advantage of ensuring real-time alerts on resource-constrained devices.

Graph representation learning (GRL) is critical for extracting insights from complex network structures, but it also raises security concerns due to potential privacy vulnerabilities in these representations. This paper investigates the structural vulnerabilities in graph neural models where sensitive topological information can be inferred through edge reconstruction attacks. Our research primarily addresses the theoretical underpinnings of cosine-similarity-based edge reconstruction attacks (COSERA), providing theoretical and empirical evidence that such attacks can perfectly reconstruct sparse Erdos Renyi graphs with independent random features as graph size increases. Conversely, we establish that sparsity is a critical factor for COSERA's effectiveness, as demonstrated through analysis and experiments on stochastic block models. Finally, we explore the resilience of (provably) private graph representations produced via noisy aggregation (NAG) mechanism against COSERA. We empirically delineate instances wherein COSERA demonstrates both efficacy and deficiency in its capacity to function as an instrument for elucidating the trade-off between privacy and utility.

Recent studies have highlighted the benefits of generating multiple synthetic datasets for supervised learning, from increased accuracy to more effective model selection and uncertainty estimation. These benefits have clear empirical support, but the theoretical understanding of them is currently very light. We seek to increase the theoretical understanding by deriving bias-variance decompositions for several settings of using multiple synthetic datasets. Our theory predicts multiple synthetic datasets to be especially beneficial for high-variance downstream predictors, and yields a simple rule of thumb to select the appropriate number of synthetic datasets in the case of mean-squared error and Brier score. We investigate how our theory works in practice by evaluating the performance of an ensemble over many synthetic datasets for several real datasets and downstream predictors. The results follow our theory, showing that our insights are also practically relevant.

Deep Learning has revolutionized the field of AI and led to remarkable achievements in applications involving image and text data. Unfortunately, there is inconclusive evidence on the merits of neural networks for structured tabular data. In this paper, we introduce a large-scale empirical study comparing neural networks against gradient-boosted decision trees on tabular data, but also transformer-based architectures against traditional multi-layer perceptrons (MLP) with residual connections. In contrast to prior work, our empirical findings indicate that neural networks are competitive against decision trees. Furthermore, we assess that transformer-based architectures do not outperform simpler variants of traditional MLP architectures on tabular datasets. As a result, this paper helps the research and practitioner communities make informed choices on deploying neural networks on future tabular data applications.

Online controlled experiments are a crucial tool to allow for confident decision-making in technology companies. A North Star metric is defined (such as long-term revenue or user retention), and system variants that statistically significantly improve on this metric in an A/B-test can be considered superior. North Star metrics are typically delayed and insensitive. As a result, the cost of experimentation is high: experiments need to run for a long time, and even then, type-II errors (i.e. false negatives) are prevalent. We propose to tackle this by learning metrics from short-term signals that directly maximise the statistical power they harness with respect to the North Star. We show that existing approaches are prone to overfitting, in that higher average metric sensitivity does not imply improved type-II errors, and propose to instead minimise the $p$-values a metric would have produced on a log of past experiments. We collect such datasets from two social media applications with over 160 million Monthly Active Users each, totalling over 153 A/B-pairs. Empirical results show that we are able to increase statistical power by up to 78% when using our learnt metrics stand-alone, and by up to 210% when used in tandem with the North Star. Alternatively, we can obtain constant statistical power at a sample size that is down to 12% of what the North Star requires, significantly reducing the cost of experimentation.

Employee's knowledge is an organization asset. Turnover may impose apparent and hidden costs and irreparable damages. To overcome and mitigate this risk, employee's condition should be monitored. Due to high complexity of analyzing well-being features, employee's turnover predicting can be delegated to machine learning techniques. In this paper, we discuss employee's attrition rate. Three different supervised learning algorithms comprising AdaBoost, SVM and RandomForest are used to benchmark employee attrition accuracy. Attained models can help out at establishing predictive analytics.

We study how to report few-shot imitation (FSI) policies' behavior errors in novel environments, a novel task named adaptable error detection (AED). The potential to cause serious damage to surrounding areas limits the application of FSI policies in real-world scenarios. Thus, a robust system is necessary to notify operators when FSI policies are inconsistent with the intent of demonstrations. We develop a cross-domain benchmark for the challenging AED task, consisting of 329 base and 158 novel environments. This task introduces three challenges, including (1) detecting behavior errors in novel environments, (2) behavior errors occurring without revealing notable changes, and (3) lacking complete temporal information of the rollout due to the necessity of online detection. To address these challenges, we propose Pattern Observer (PrObe) to parse discernible patterns in the policy feature representations of normal or error states, whose effectiveness is verified in the proposed benchmark. Through our comprehensive evaluation, PrObe consistently surpasses strong baselines and demonstrates a robust capability to identify errors arising from a wide range of FSI policies. Moreover, we conduct comprehensive ablations and experiments (error correction, demonstration quality, etc.) to validate the practicality of our proposed task and methodology.