There is a growing need for investigating how machine learning models operate. With this work, we aim to understand trained machine learning models by questioning their data preferences. We propose a mathematical framework that allows us to probe trained models and identify their preferred samples in various scenarios including prediction-risky, parameter-sensitive, or model-contrastive samples. To showcase our framework, we pose these queries to a range of models trained on a range of classification and regression tasks, and receive answers in the form of generated data.

Large-scale surveys are essential tools for informing social science research and policy, but running surveys is costly and time-intensive. If we could accurately simulate group-level survey results, this would therefore be very valuable to social science research. Prior work has explored the use of large language models (LLMs) for simulating human behaviors, mostly through prompting. In this paper, we are the first to specialize LLMs for the task of simulating survey response distributions. As a testbed, we use country-level results from two global cultural surveys. We devise a fine-tuning method based on first-token probabilities to minimize divergence between predicted and actual response distributions for a given question. Then, we show that this method substantially outperforms other methods and zero-shot classifiers, even on unseen questions, countries, and a completely unseen survey. While even our best models struggle with the task, especially on unseen questions, our results demonstrate the benefits of specialization for simulation, which may accelerate progress towards sufficiently accurate simulation in the future.

We introduce Social Bayesian Optimization (SBO), a vote-efficient algorithm for consensus-building in collective decision-making. In contrast to single-agent scenarios, collective decision-making encompasses group dynamics that may distort agents' preference feedback, thereby impeding their capacity to achieve a social-influence-free consensus -- the most preferable decision based on the aggregated agent utilities. We demonstrate that under mild rationality axioms, reaching social-influence-free consensus using noisy feedback alone is impossible. To address this, SBO employs a dual voting system: cheap but noisy public votes (e.g., show of hands in a meeting), and more accurate, though expensive, private votes (e.g., one-to-one interview). We model social influence using an unknown social graph and leverage the dual voting system to efficiently learn this graph. Our theoretical findigns show that social graph estimation converges faster than the black-box estimation of agents' utilities, allowing us to reduce reliance on costly private votes early in the process. This enables efficient consensus-building primarily through noisy public votes, which are debiased based on the estimated social graph to infer social-influence-free feedback. We validate the efficacy of SBO across multiple real-world applications, including thermal comfort, team building, travel negotiation, and energy trading collaboration.

Phishing attacks remain a significant threat in the digital age, yet organizations lack effective methods to tackle phishing attacks without leaking sensitive information. Phish bowl initiatives are a vital part of cybersecurity efforts against these attacks. However, traditional phish bowls require manual anonymization and are often limited to internal use. To overcome these limitations, we introduce AdaPhish, an AI-powered phish bowl platform that automatically anonymizes and analyzes phishing emails using large language models (LLMs) and vector databases. AdaPhish achieves real-time detection and adaptation to new phishing tactics while enabling long-term tracking of phishing trends. Through automated reporting, adaptive analysis, and real-time alerts, AdaPhish presents a scalable, collaborative solution for phishing detection and cybersecurity education.

Conformal prediction methods create prediction bands with distribution-free guarantees but do not explicitly capture epistemic uncertainty, which can lead to overconfident predictions in data-sparse regions. Although recent conformal scores have been developed to address this limitation, they are typically designed for specific tasks, such as regression or quantile regression. Moreover, they rely on particular modeling choices for epistemic uncertainty, restricting their applicability. We introduce $\texttt{EPICSCORE}$, a model-agnostic approach that enhances any conformal score by explicitly integrating epistemic uncertainty. Leveraging Bayesian techniques such as Gaussian Processes, Monte Carlo Dropout, or Bayesian Additive Regression Trees, $\texttt{EPICSCORE}$ adaptively expands predictive intervals in regions with limited data while maintaining compact intervals where data is abundant. As with any conformal method, it preserves finite-sample marginal coverage. Additionally, it also achieves asymptotic conditional coverage. Experiments demonstrate its good performance compared to existing methods. Designed for compatibility with any Bayesian model, but equipped with distribution-free guarantees, $\texttt{EPICSCORE}$ provides a general-purpose framework for uncertainty quantification in prediction problems.

In decentralized multi-agent deep reinforcement learning (MADRL), communication can help agents to gain a better understanding of the environment to better coordinate their behaviors. Nevertheless, communication may involve uncertainty, which potentially introduces variance to the learning of decentralized agents. In this paper, we focus on a specific decentralized MADRL setting with communication and conduct a theoretical analysis to study the variance that is caused by communication in policy gradients. We propose modular techniques to reduce the variance in policy gradients during training. We adopt our modular techniques into two existing algorithms for decentralized MADRL with communication and evaluate them on multiple tasks in the StarCraft Multi-Agent Challenge and Traffic Junction domains. The results show that decentralized MADRL communication methods extended with our proposed techniques not only achieve high-performing agents but also reduce variance in policy gradients during training.

For very large values of $k$, we consider methods for fast $k$-means clustering of massive datasets with $10^7\sim10^9$ points in high-dimensions ($d\geq100$). All current practical methods for this problem have runtimes at least $\Omega(k^2)$. We find that initialization routines are not a bottleneck for this case. Instead, it is critical to improve the speed of Lloyd's local-search algorithm, particularly the step that reassigns points to their closest center. Attempting to improve this step naturally leads us to leverage approximate nearest-neighbor search methods, although this alone is not enough to be practical. Instead, we propose a family of problems we call "Seeded Approximate Nearest-Neighbor Search", for which we propose "Seeded Search-Graph" methods as a solution.

Electroencephalogram (EEG) provides a non-invasive, highly accessible, and cost-effective solution for Alzheimer's Disease (AD) detection. However, existing methods, whether based on manual feature extraction or deep learning, face two major challenges: the lack of large-scale datasets for robust feature learning and evaluation, and poor detection performance due to inter-subject variations. To address these challenges, we curate an EEG-AD corpus containing 813 subjects, which forms the world's largest EEG-AD dataset to the best of our knowledge. Using this unique dataset, we propose LEAD, the first large foundation model for EEG-based AD detection. Our method encompasses an entire pipeline, from data selection and preprocessing to self-supervised contrastive pretraining, fine-tuning, and key setups such as subject-independent evaluation and majority voting for subject-level detection. We pre-train the model on 11 EEG datasets and unified fine-tune it on 5 AD datasets. Our self-supervised pre-training design includes sample-level and subject-level contrasting to extract useful general EEG features. Fine-tuning is performed on 5 channel-aligned datasets together. The backbone encoder incorporates temporal and channel embeddings to capture features across both temporal and spatial dimensions. Our method demonstrates outstanding AD detection performance, achieving up to a 9.86% increase in F1 score at the sample-level and up to a 9.31% at the subject-level compared to state-of-the-art methods. The results of our model strongly confirm the effectiveness of contrastive pre-training and channel-aligned unified fine-tuning for addressing inter-subject variation. The source code is at https://github.com/DL4mHealth/LEAD.

We examine the representation of African American English (AAE) in large language models (LLMs), exploring (a) the perceptions Black Americans have of how effective these technologies are at producing authentic AAE, and (b) in what contexts Black Americans find this desirable. Through both a survey of Black Americans ($n=$ 104) and annotation of LLM-produced AAE by Black Americans ($n=$ 228), we find that Black Americans favor choice and autonomy in determining when AAE is appropriate in LLM output. They tend to prefer that LLMs default to communicating in Mainstream U.S. English in formal settings, with greater interest in AAE production in less formal settings. When LLMs were appropriately prompted and provided in context examples, our participants found their outputs to have a level of AAE authenticity on par with transcripts of Black American speech. Select code and data for our project can be found here: https://github.com/smelliecat/AAEMime.git

We propose a new long-context financial benchmark, FailSafeQA, designed to test the robustness and context-awareness of LLMs against six variations in human-interface interactions in LLM-based query-answer systems within finance. We concentrate on two case studies: Query Failure and Context Failure. In the Query Failure scenario, we perturb the original query to vary in domain expertise, completeness, and linguistic accuracy. In the Context Failure case, we simulate the uploads of degraded, irrelevant, and empty documents. We employ the LLM-as-a-Judge methodology with Qwen2.5-72B-Instruct and use fine-grained rating criteria to define and calculate Robustness, Context Grounding, and Compliance scores for 24 off-the-shelf models. The results suggest that although some models excel at mitigating input perturbations, they must balance robust answering with the ability to refrain from hallucinating. Notably, Palmyra-Fin-128k-Instruct, recognized as the most compliant model, maintained strong baseline performance but encountered challenges in sustaining robust predictions in 17% of test cases. On the other hand, the most robust model, OpenAI o3-mini, fabricated information in 41% of tested cases. The results demonstrate that even high-performing models have significant room for improvement and highlight the role of FailSafeQA as a tool for developing LLMs optimized for dependability in financial applications. The dataset is available at: https://huggingface.co/datasets/Writer/FailSafeQA