Progress in computer use agents (CUAs) has been constrained by the absence of large and high-quality datasets that capture how humans interact with a computer. While LLMs have thrived on abundant textual data, no comparable corpus exists for CUA trajectories. To address these gaps, we introduce FaraGen, a novel synthetic data generation system for multi-step web tasks. FaraGen can propose diverse tasks from frequently used websites, generate multiple solution attempts, and filter successful trajectories using multiple verifiers. It achieves high throughput, yield, and diversity for multi-step web tasks, producing verified trajectories at approximately $1 each. We use this data to train Fara-7B, a native CUA model that perceives the computer using only screenshots, executes actions via predicted coordinates, and is small enough to run on-device. We find that Fara-7B outperforms other CUA models of comparable size on benchmarks like WebVoyager, Online-Mind2Web, and WebTailBench -- our novel benchmark that better captures under-represented web tasks in pre-existing benchmarks. Furthermore, Fara-7B is competitive with much larger frontier models, illustrating key benefits of scalable data generation systems in advancing small efficient agentic models. We are making Fara-7B open-weight on Microsoft Foundry and HuggingFace, and we are releasing WebTailBench.

The ever-increasing number of Android devices and the accelerated evolution of malware, reaching over 35 million samples by 2024, highlight the critical importance of effective detection methods. Attackers are now using Artificial Intelligence to create sophisticated malware variations that can easily evade traditional detection techniques. Although machine learning has shown promise in malware classification, its success relies heavily on the availability of up-to-date, high-quality datasets. The scarcity and high cost of obtaining and labeling real malware samples presents significant challenges in developing robust detection models. In this paper, we propose MalSynGen, a Malware Synthetic Data Generation methodology that uses a conditional Generative Adversarial Network (cGAN) to generate synthetic tabular data. This data preserves the statistical properties of real-world data and improves the performance of Android malware classifiers. We evaluated the effectiveness of this approach using various datasets and metrics that assess the fidelity of the generated data, its utility in classification, and the computational efficiency of the process. Our experiments demonstrate that MalSynGen can generalize across different datasets, providing a viable solution to address the issues of obsolescence and low quality data in malware detection. With approximately 3 billion Android devices in operation worldwide [1], the mobile cybersecurity landscape faces formidable challenges. In 2024 alone, Kaspersky reported over 33.3 million cyberattacks targeting smartphone users globally, encompassing diverse forms of malware and unwanted software [2]. Adding to this problem, attackers are using Artificial Intelligence (AI) to rapidly generate new malware variants by exploiting patterns learned from existing malware [3].

The rapid advancement of generators (e.g., StyleGAN, Midjourney, DALL-E) has produced highly realistic synthetic images, posing significant challenges to digital media authenticity. These generators are typically based on a few core architectural families, primarily Generative Adversarial Networks (GANs) and Diffusion Models (DMs). A critical vulnerability in current forensics is the failure of detectors to achieve cross-generator generalization, especially when crossing architectural boundaries (e.g., from GANs to DMs). We hypothesize that this gap stems from fundamental differences in the artifacts produced by these \textbf{distinct architectures}. In this work, we provide a theoretical analysis explaining how the distinct optimization objectives of the GAN and DM architectures lead to different manifold coverage behaviors. We demonstrate that GANs permit partial coverage, often leading to boundary artifacts, while DMs enforce complete coverage, resulting in over-smoothing patterns. Motivated by this analysis, we propose the \textbf{Tri}archy \textbf{Detect}or (TriDetect), a semi-supervised approach that enhances binary classification by discovering latent architectural patterns within the "fake" class. TriDetect employs balanced cluster assignment via the Sinkhorn-Knopp algorithm and a cross-view consistency mechanism, encouraging the model to learn fundamental architectural distincts. We evaluate our approach on two standard benchmarks and three in-the-wild datasets against 13 baselines to demonstrate its generalization capability to unseen generators.

Approximating training-point influence on test predictions is critical for deploying deep-learning vision models, essential for locating noisy data. Though the influence function was proposed for attributing how infinitesimal up-weighting or removal of individual training examples affects model outputs, its implementation is still challenging in deep-learning vision models: inverse-curvature computations are expensive, and training non-stationarity invalidates static approximations. Prior works use iterative solvers and low-rank surrogates to reduce cost, but offline computation lags behind training dynamics, and missing confidence calibration yields fragile rankings that misidentify critical examples. To address these challenges, we introduce a Stability-Guided Online Influence Framework (SG-OIF), the first framework that treats algorithmic stability as a real-time controller, which (i) maintains lightweight anchor IHVPs via stochastic Richardson and preconditioned Neumann; (ii) proposes modular curvature backends to modulate per-example influence scores using stability-guided residual thresholds, anomaly gating, and confidence. Experimental results show that SG-OIF achieves SOTA (State-Of-The-Art) on noise-label and out-of-distribution detection tasks across multiple datasets with various corruption. Notably, our approach achieves 91.1\% accuracy in the top 1\% prediction samples on the CIFAR-10 (20\% asym), and gets 99.8\% AUPR score on MNIST, effectively demonstrating that this framework is a practical controller for online influence estimation.

Large Language Models (LLMs) are increasingly deployed in high-stakes domains such as science, law, and healthcare, where accurate expressions of uncertainty are essential for reliability and trust. However, current LLMs are often observed to generate incorrect answers with high confidence, a phenomenon known as "overconfidence". Recent efforts have focused on calibrating LLMs' verbalized confidence: i.e., their expressions of confidence in text form, such as "I am 80% confident that...". Existing approaches either rely on prompt engineering or fine-tuning with heuristically generated uncertainty estimates, both of which have limited effectiveness and generalizability. Motivated by the notion of proper scoring rules for calibration in classical machine learning models, we introduce ConfTuner, a simple and efficient fine-tuning method that introduces minimal overhead and does not require ground-truth confidence scores or proxy confidence estimates. ConfTuner relies on a new loss function, tokenized Brier score, which we theoretically prove to be a proper scoring rule, intuitively meaning that it "correctly incentivizes the model to report its true probability of being correct". ConfTuner improves calibration across diverse reasoning tasks and generalizes to black-box models such as GPT-4o. Our results further show that better-calibrated confidence enables downstream gains in self-correction and model cascade, advancing the development of trustworthy LLM systems. The code is available at https://github.com/liushiliushi/ConfTuner.

Research on backdoor attacks in Federated Learning (FL) has accelerated in recent years, with new attacks and defenses continually proposed in an escalating arms race. However, the evaluation of these methods remains neither standardized nor reliable. First, there are severe inconsistencies in the evaluation settings across studies, and many rely on unrealistic threat models. Second, our code review uncovers semantic bugs in the official codebases of several attacks that artificially inflate their reported performance. These issues raise fundamental questions about whether current methods are truly effective or simply overfitted to narrow experimental setups. We introduce \textbf{BackFed}, a benchmark designed to standardize and stress-test FL backdoor evaluation by unifying attacks and defenses under a common evaluation framework that mirrors realistic FL deployments. Our benchmark on three representative datasets with three distinct architectures reveals critical limitations of existing methods. Malicious clients often require excessive training time and computation, making them vulnerable to server-enforced time constraints. Meanwhile, several defenses incur severe accuracy degradation or aggregation overhead. Popular defenses and attacks achieve limited performance in our benchmark, which challenges their previous efficacy claims. We establish BackFed as a rigorous and fair evaluation framework that enables more reliable progress in FL backdoor research.

Abstract--Time series anomaly detection is widely used in IoT and cyber-physical systems, yet its evaluation remains challenging due to diverse application objectives and heterogeneous metric assumptions. This study introduces a problem-oriented framework that reinterprets existing metrics based on the specific evaluation challenges they are designed to address, rather than their mathematical forms or output structures. We categorize over twenty commonly used metrics into six dimensions: (1) basic accuracy-driven evaluation, (2) timeliness-aware reward mechanisms, (3) tolerance to labeling imprecision, (4) penalties reflecting human-audit cost, (5) robustness against random or inflated scores, and (6) parameter-free comparability for cross-dataset benchmark-ing. Comprehensive experiments are conducted to examine metric behavior under genuine, random, and oracle detection scenarios. By comparing their resulting score distributions, we quantify each metric's discriminative ability--its capability to distinguish meaningful detections from random noise. The results show that while most event-level metrics exhibit strong separability, several widely used metrics (e.g., NAB, Point-Adjust) demonstrate limited resistance to random-score inflation. These findings reveal that metric suitability must be inherently task-dependent and aligned with the operational objectives of IoT applications. The proposed framework offers a unified analytical perspective for understanding existing metrics and provides practical guidance for selecting or developing more context-aware, robust, and fair evaluation methodologies for time series anomaly detection. He emergence of the Internet of Things (IoT) has accelerated digital transformation across numerous domains. Its defining characteristic lies in the large-scale deployment of intelligent and heterogeneous devices--such as sensors, actuators, and RFID systems--that are interconnected via the Internet to enable autonomous communication without human intervention [1]. Currently, more than 12 billion IoT devices are in operation, and this number is projected to reach 125 billion by 2030 [2]. Consequently, the volume of data generated by these devices continues to soar, with an expected total of 79.4 ZB by 2025 [3]. In industrial contexts, the integration of IoT technologies has driven the ongoing Industry 4.0 revolution, emphasizing connectivity, automation, and intelligence. Kaixiang Y ang, Jiarong Liu, Y upeng Song, and Y ujue Zhou are with the School of Artificial Intelligence, Y unnan University, Kunming 650091, China. Shuanghua Y ang is with Beijing Normal University - Hong Kong Baptist University, Zhuhai 519087, China. This work was supported in part by the Y unnan Fundamental Research Projects under Grant 202401AU070151, and in part by the Y unnan Provincial Science and Technology Talent and Platform Plan under Grant 202505AF350053.

Sampling multiple outputs from a Large Language Model (LLM) and selecting the most frequent (Self-consistency) or highest-scoring (Best-of-N) candidate is a popular approach to achieve higher accuracy in tasks with discrete final answers. Best-of-N (BoN) selects the output with the highest reward, and with perfect rewards, it often achieves near-perfect accuracy. With imperfect rewards from reward models, however, BoN fails to reliably find the correct answer and its performance degrades drastically. We consider the distribution of BoN's outputs and highlight that, although the correct answer does not usually have a probability close to one under imperfect rewards, it is often the most likely outcome. This suggests that the mode of this distribution can be more reliably correct than a sample from it. Based on this idea, we propose Majority-of-the-Bests (MoB), a novel selection mechanism that estimates the output distribution of BoN via bootstrapping and selects its mode. Experimental results across five benchmarks, three different base LLMs, and two reward models demonstrate consistent improvements over BoN in 25 out of 30 setups. We also provide theoretical results for the consistency of the bootstrapping. MoB serves as a simple, yet strong alternative to BoN and self-consistency, and more broadly, motivates further research in more nuanced selection mechanisms.

Ecological Momentary Assessment provides real-time data on suicidal thoughts and behaviors, but predicting suicide attempts remains challenging due to their rarity and patient heterogeneity. We show that single models fit to all patients perform poorly, while individualized models improve performance but still overfit to patients with limited data. To address this, we introduce Latent Similarity Gaussian Processes (LSGPs) to capture patient heterogeneity, enabling those with little data to leverage similar patients' trends. Preliminary results show promise: even without kernel-design, we outperform all but one baseline while offering a new understanding of patient similarity.

Clinical and genomic models are both used to predict breast cancer outcomes, but they are often combined using simple linear rules that do not account for how their risk scores relate, especially at the extremes. Using the METABRIC breast cancer cohort, we studied whether directly modeling the joint relationship between clinical and genomic machine learning risk scores could improve risk stratification for 5-year cancer-specific mortality. We created a binary 5-year cancer-death outcome and defined two sets of predictors: a clinical set (demographic, tumor, and treatment variables) and a genomic set (gene-expression $z$-scores). We trained several supervised classifiers, such as Random Forest and XGBoost, and used 5-fold cross-validated predicted probabilities as unbiased risk scores. These scores were converted to pseudo-observations on $(0,1)^2$ to fit Gaussian, Clayton, and Gumbel copulas. Clinical models showed good discrimination (AUC 0.783), while genomic models had moderate performance (AUC 0.681). The joint distribution was best captured by a Gaussian copula (bootstrap $p=0.997$), which suggests a symmetric, moderately strong positive relationship. When we grouped patients based on this relationship, Kaplan-Meier curves showed clear differences: patients who were high-risk in both clinical and genomic scores had much poorer survival than those high-risk in only one set. These results show that copula-based fusion works in real-world cohorts and that considering dependencies between scores can better identify patient subgroups with the worst prognosis.