This dissertation establishes the contexture theory to mathematically characterize the mechanism of representation learning, or pretraining. Despite the remarkable empirical success of foundation models, it is not very clear what representations they learn, and why these representations are useful for various downstream tasks. A scientific understanding of representation learning is critical, especially at this point when scaling up the model size is producing diminishing returns, and designing new pretraining methods is imperative for further progress. Prior work treated different representation learning methods quite differently, whereas the contexture theory provides a unified framework for analyzing these methods. The central argument is that a representation is learned from the association between the input X and a context variable A. We prove that if an encoder captures the maximum information of this association, in which case we say that the encoder learns the contexture, then it will be optimal on the class of tasks that are compatible with the context. We also show that a context is the most useful when the association between X and A is neither too strong nor too weak. The important implication of the contexture theory is that increasing the model size alone will achieve diminishing returns, and further advancements require better contexts. We demonstrate that many pretraining objectives can learn the contexture, including supervised learning, self-supervised learning, generative models, etc. Then, we introduce two general objectives -- SVME and KISE, for learning the contexture. We also show how to mix multiple contexts together, an effortless way to create better contexts from existing ones. Then, we prove statistical learning bounds for representation learning. Finally, we discuss the effect of the data distribution shift from pretraining to the downstream task.

Empirical inference attacks are a popular approach for evaluating the privacy risk of data release mechanisms in practice. While an active attack literature exists to evaluate machine learning models or synthetic data release, we currently lack comparable methods for fixed aggregate statistics, in particular when only a limited number of statistics are released. We here propose an inference attack framework against fixed aggregate statistics and an attribute inference attack called DeSIA. We instantiate DeSIA against the U.S. Census PPMF dataset and show it to strongly outperform reconstruction-based attacks. In particular, we show DeSIA to be highly effective at identifying vulnerable users, achieving a true positive rate of 0.14 at a false positive rate of $10^{-3}$. We then show DeSIA to perform well against users whose attributes cannot be verified and when varying the number of aggregate statistics and level of noise addition. We also perform an extensive ablation study of DeSIA and show how DeSIA can be successfully adapted to the membership inference task. Overall, our results show that aggregation alone is not sufficient to protect privacy, even when a relatively small number of aggregates are being released, and emphasize the need for formal privacy mechanisms and testing before aggregate statistics are released.

Geomagnetic storms are large-scale disturbances of the Earth's magnetosphere driven by solar wind interactions, posing significant risks to space-based and ground-based infrastructure. The Disturbance Storm Time (Dst) index quantifies geomagnetic storm intensity by measuring global magnetic field variations. This study applies symbolic regression to derive data-driven equations describing the temporal evolution of the Dst index. We use historical data from the NASA OMNIweb database, including solar wind density, bulk velocity, convective electric field, dynamic pressure, and magnetic pressure. The PySR framework, an evolutionary algorithm-based symbolic regression library, is used to identify mathematical expressions linking dDst/dt to key solar wind. The resulting models include a hierarchy of complexity levels and enable a comparison with well-established empirical models such as the Burton-McPherron-Russell and O'Brien-McPherron models. The best-performing symbolic regression models demonstrate superior accuracy in most cases, particularly during moderate geomagnetic storms, while maintaining physical interpretability. Performance evaluation on historical storm events includes the 2003 Halloween Storm, the 2015 St. Patrick's Day Storm, and a 2017 moderate storm. The results provide interpretable, closed-form expressions that capture nonlinear dependencies and thresholding effects in Dst evolution.

Should a large language model (LLM) be used as a therapist? In this paper, we investigate the use of LLMs to *replace* mental health providers, a use case promoted in the tech startup and research space. We conduct a mapping review of therapy guides used by major medical institutions to identify crucial aspects of therapeutic relationships, such as the importance of a therapeutic alliance between therapist and client. We then assess the ability of LLMs to reproduce and adhere to these aspects of therapeutic relationships by conducting several experiments investigating the responses of current LLMs, such as `gpt-4o`. Contrary to best practices in the medical community, LLMs 1) express stigma toward those with mental health conditions and 2) respond inappropriately to certain common (and critical) conditions in naturalistic therapy settings -- e.g., LLMs encourage clients' delusional thinking, likely due to their sycophancy. This occurs even with larger and newer LLMs, indicating that current safety practices may not address these gaps. Furthermore, we note foundational and practical barriers to the adoption of LLMs as therapists, such as that a therapeutic alliance requires human characteristics (e.g., identity and stakes). For these reasons, we conclude that LLMs should not replace therapists, and we discuss alternative roles for LLMs in clinical therapy.

Length of hospital stay is a critical metric for assessing healthcare quality and optimizing hospital resource management. This study aims to identify factors influencing LoS within the Italian healthcare context, using a dataset of hospitalization records from over 60 healthcare facilities in the Piedmont region, spanning from 2020 to 2023. We explored a variety of features, including patient characteristics, comorbidities, admission details, and hospital-specific factors. Significant correlations were found between LoS and features such as age group, comorbidity score, admission type, and the month of admission. Machine learning models, specifically CatBoost and Random Forest, were used to predict LoS. The highest R2 score, 0.49, was achieved with CatBoost, demonstrating good predictive performance.

Abstract--Large Language Models (LLMs) are increasingly used as automated judges for evaluating text quality, code c or-rectness, and argument strength. However, these LLM-as-a-judge systems are vulnerable to adversarial attacks that can mani pulate their assessments. This paper investigates the vulnerabil ity of LLM-as-a-judge systems to prompt injection attacks, drawi ng insights from both academic literature and practical solutio ns from the "LLMs: Y ou Can't Please Them All" Kaggle competition. We present a comprehensive framework for developing and evaluating adversarial attacks against LLM judges, distinguis hing between content-author attacks and system-prompt attacks . Through rigorous statistical a nalysis (n=50 prompts per condition, bootstrap confidence interval s), we demonstrate that sophisticated attacks can achieve succes s rates of up to 73.8% against popular LLM judges, with Contextual Misdirection being the most effective method against Gemma models at 67.7%. We find that smaller models like Gemma-3-4B-Instruct are more vulnerable (65.9% average success rat e) than their larger counterparts, and that attacks show high transferability (50.5-62.6%) We compare our approach with recent work including Universal-Prompt-Injection [1] and AdvPrompter [2], demonstrating b oth complementary insights and novel contributions. Our findin gs highlight critical vulnerabilities in current LLM-as-a-j udge systems and provide recommendations for developing more robus t evaluation frameworks, including using multi-model commi ttees with diverse architectures and preferring comparative ass essment over absolute scoring methods. T o ensure reproducibility, we release our code, evaluation harness, and processed datase ts.

Accurate prediction of lithium-ion battery lifespan is vital for ensuring operational reliability and reducing maintenance costs in applications like electric vehicles and smart grids. This study presents a hybrid learning framework for precise battery lifespan prediction, integrating dynamic multi-source data fusion with a stacked ensemble (SE) modeling approach. By leveraging heterogeneous datasets from the National Aeronautics and Space Administration (NASA), Center for Advanced Life Cycle Engineering (CALCE), MIT-Stanford-Toyota Research Institute (TRC), and nickel cobalt aluminum (NCA) chemistries, an entropy-based dynamic weighting mechanism mitigates variability across heterogeneous datasets. The SE model combines Ridge regression, long short-term memory (LSTM) networks, and eXtreme Gradient Boosting (XGBoost), effectively capturing temporal dependencies and nonlinear degradation patterns. It achieves a mean absolute error (MAE) of 0.0058, root mean square error (RMSE) of 0.0092, and coefficient of determination (R2) of 0.9839, outperforming established baseline models with a 46.2% improvement in R2 and an 83.2% reduction in RMSE. Shapley additive explanations (SHAP) analysis identifies differential discharge capacity (Qdlin) and temperature of measurement (Temp_m) as critical aging indicators. This scalable, interpretable framework enhances battery health management, supporting optimized maintenance and safety across diverse energy storage systems, thereby contributing to improved battery health management in energy storage systems.

Using LLMs in healthcare, Computer-Supported Cooperative Work, and Social Computing requires the examination of ethical and social norms to ensure safe incorporation into human life. We conducted a mixed-method study, including an online survey with 111 participants and an interview study with 38 experts, to investigate the AI ethics and social norms in ChatGPT as everyday life tools. This study aims to evaluate whether ChatGPT in an empirical context operates following ethics and social norms, which is critical for understanding actions in industrial and academic research and achieving machine ethics. The findings of this study provide initial insights into six important aspects of AI ethics, including bias, trustworthiness, security, toxicology, social norms, and ethical data. Significant obstacles related to transparency and bias in unsupervised data collection methods are identified as ChatGPT's ethical concerns.

Efforts to ensure the safety of large language models (LLMs) include safety fine-tuning, evaluation, and red teaming. However, despite the widespread use of the Retrieval-Augmented Generation (RAG) framework, AI safety work focuses on standard LLMs, which means we know little about how RAG use cases change a model's safety profile. We conduct a detailed comparative analysis of RAG and non-RAG frameworks with eleven LLMs. We find that RAG can make models less safe and change their safety profile. We explore the causes of this change and find that even combinations of safe models with safe documents can cause unsafe generations. In addition, we evaluate some existing red teaming methods for RAG settings and show that they are less effective than when used for non-RAG settings. Our work highlights the need for safety research and red-teaming methods specifically tailored for RAG LLMs.

Drug discovery remains a formidable challenge: more than 90 percent of candidate molecules fail in clinical evaluation, and development costs often exceed one billion dollars per approved therapy. Disparate data streams, from genomics and transcriptomics to chemical libraries and clinical records, hinder coherent mechanistic insight and slow progress. Meanwhile, large language models excel at reasoning and tool integration but lack the modular specialization and iterative memory required for regulated, hypothesis-driven workflows. We introduce PharmaSwarm, a unified multi-agent framework that orchestrates specialized LLM "agents" to propose, validate, and refine hypotheses for novel drug targets and lead compounds. Each agent accesses dedicated functionality--automated genomic and expression analysis; a curated biomedical knowledge graph; pathway enrichment and network simulation; interpretable binding affinity prediction--while a central Evaluator LLM continuously ranks proposals by biological plausibility, novelty, in silico efficacy, and safety. A shared memory layer captures validated insights and fine-tunes underlying submodels over time, yielding a self-improving system. Deployable on low-code platforms or Kubernetes-based microservices, PharmaSwarm supports literature-driven discovery, omics-guided target identification, and market-informed repurposing. We also describe a rigorous four-tier validation pipeline spanning retrospective benchmarking, independent computational assays, experimental testing, and expert user studies to ensure transparency, reproducibility, and real-world impact. By acting as an AI copilot, PharmaSwarm can accelerate translational research and deliver high-confidence hypotheses more efficiently than traditional pipelines.