Large language models are rapidly transforming social science research by enabling the automation of labor-intensive tasks like data annotation and text analysis. However, LLM outputs vary significantly depending on the implementation choices made by researchers (e.g., model selection or prompting strategy). Such variation can introduce systematic biases and random errors, which propagate to downstream analyses and cause Type I (false positive), Type II (false negative), Type S (wrong sign), or Type M (exaggerated effect) errors. We call this phenomenon where configuration choices lead to incorrect conclusions LLM hacking. We find that intentional LLM hacking is strikingly simple. By replicating 37 data annotation tasks from 21 published social science studies, we show that, with just a handful of prompt paraphrases, virtually anything can be presented as statistically significant. Beyond intentional manipulation, our analysis of 13 million labels from 18 different LLMs across 2361 realistic hypotheses shows that there is also a high risk of accidental LLM hacking, even when following standard research practices. We find incorrect conclusions in approximately 31% of hypotheses for state-of-the-art LLMs, and in half the hypotheses for smaller language models. While higher task performance and stronger general model capabilities reduce LLM hacking risk, even highly accurate models remain susceptible. The risk of LLM hacking decreases as effect sizes increase, indicating the need for more rigorous verification of LLM-based findings near significance thresholds. We analyze 21 mitigation techniques and find that human annotations provide crucial protection against false positives. Common regression estimator correction techniques can restore valid inference but trade off Type I vs. Type II errors. We publish a list of practical recommendations to prevent LLM hacking.

Stroke is one of the leading causes of death globally, making early and accurate diagnosis essential for improving patient outcomes, particularly in emergency settings where timely intervention is critical. CT scans are the key imaging modality because of their speed, accessibility, and cost-effectiveness. This study proposed an artificial intelligence framework for multiclass stroke classification (ischemic, hemorrhagic, and no stroke) using CT scan images from a dataset provided by the Republic of Turkey's Ministry of Health. The proposed method adopted MaxViT, a state-of-the-art Vision Transformer, as the primary deep learning model for image-based stroke classification, with additional transformer variants (vision transformer, transformer-in-transformer, and ConvNext). To enhance model generalization and address class imbalance, we applied data augmentation techniques, including synthetic image generation. The MaxViT model trained with augmentation achieved the best performance, reaching an accuracy and F1-score of 98.00%, outperforming all other evaluated models and the baseline methods. The primary goal of this study was to distinguish between stroke types with high accuracy while addressing crucial issues of transparency and trust in artificial intelligence models. To achieve this, Explainable Artificial Intelligence (XAI) was integrated into the framework, particularly Grad-CAM++. It provides visual explanations of the model's decisions by highlighting relevant stroke regions in the CT scans and establishing an accurate, interpretable, and clinically applicable solution for early stroke detection. This research contributed to the development of a trustworthy AI-assisted diagnostic tool for stroke, facilitating its integration into clinical practice and enhancing access to timely and optimal stroke diagnosis in emergency departments, thereby saving more lives.

We study the problem active sequential hypothesis testing, also known as pure exploration: given a new task, the learner adaptively collects data from the environment to efficiently determine an underlying correct hypothesis. A classical instance of this problem is the task of identifying the best arm in a multi-armed bandit problem (a.k.a. BAI, Best-Arm Identification), where actions index hypotheses. Another important case is generalized search, a problem of determining the correct label through a sequence of strategically selected queries that indirectly reveal information about the label. In this work, we introduce In-Context Pure Exploration (ICPE), which meta-trains Transformers to map observation histories to query actions and a predicted hypothesis, yielding a model that transfers in-context. At inference time, ICPE actively gathers evidence on new tasks and infers the true hypothesis without parameter updates. Across deterministic, stochastic, and structured benchmarks, including BAI and generalized search, ICPE is competitive with adaptive baselines while requiring no explicit modeling of information structure. Our results support Transformers as practical architectures for general sequential testing.

Software is an essential component of research. However, little attention has been paid to it compared with that paid to research data. Recently, there has been an increase in efforts to acknowledge and highlight the importance of software in research activities. Structured metadata from platforms like bio.tools, Bioconductor, and Galaxy ToolShed offers valuable insights into research software in the Life Sciences. Although originally intended to support discovery and integration, this metadata can be repurposed for large-scale analysis of software practices. However, its quality and completeness vary across platforms, reflecting diverse documentation practices. To gain a comprehensive view of software development and sustainability, consolidating this metadata is necessary, but requires robust mechanisms to address its heterogeneity and scale. This article presents an evaluation of instruction-tuned large language models for the task of software metadata identity resolution, a critical step in assembling a cohesive collection of research software. Such a collection is the reference component for the Software Observatory at OpenEBench, a platform that aggregates metadata to monitor the FAIRness of research software in the Life Sciences. We benchmarked multiple models against a human-annotated gold standard, examined their behavior on ambiguous cases, and introduced an agreement-based proxy for high-confidence automated decisions. The proxy achieved high precision and statistical robustness, while also highlighting the limitations of current models and the broader challenges of automating semantic judgment in FAIR-aligned software metadata across registries and repositories.

Neural networks trained on real-world data often exhibit biases while simultaneously being vulnerable to privacy attacks aimed at extracting sensitive information. Despite extensive research on each problem individually, their intersection remains poorly understood. In this work, we investigate the privacy impact of spurious correlation bias. We introduce \emph{spurious privacy leakage}, a phenomenon in which spurious groups are significantly more vulnerable to privacy attacks than non-spurious groups. We observe that privacy disparity between groups increases in tasks with simpler objectives (e.g. fewer classes) due to spurious features. Counterintuitively, we demonstrate that spurious robust methods, designed to reduce spurious bias, fail to mitigate privacy disparity. Our analysis reveals that this occurs because robust methods can reduce reliance on spurious features for prediction, but do not prevent their memorization during training. Finally, we systematically compare the privacy of different model architectures trained with spurious data, demonstrating that, contrary to previous work, architectural choice can affect privacy evaluation.

T cell receptor (TCR) recognition of peptide-MHC (pMHC) complexes is a central component of adaptive immunity, with implications for vaccine design, cancer immunotherapy, and autoimmune disease. While recent advances in machine learning have improved prediction of TCR-pMHC binding, the most effective approaches are black-box transformer models that cannot provide a rationale for predictions. Post-hoc explanation methods can provide insight with respect to the input but do not explicitly model biochemical mechanisms (e.g. "Explain-by-design" models (i.e., with architectural components that can be examined directly after training) have been explored in other domains, but have not been used for TCR-pMHC binding. We propose explainable model layers (TCR-EML) that can be incorporated into protein-language model backbones for TCR-pMHC modeling. Our approach uses prototype layers for amino acid residue contacts drawn from known TCR-pMHC binding mechanisms, enabling high-quality explanations for predicted TCR-pMHC binding. Experiments of our proposed method on large-scale datasets demonstrate competitive predictive accuracy and generalization, and evaluation on the TCR-XAI benchmark demonstrates improved explainability compared with existing approaches. For the adaptive immune system, T cells are essential for detecting and responding to antigens from pathogens such as viruses, bacteria, and cancer cells (Joglekar & Li, 2021), as well as in autoimmune contexts.

Many high-stakes AI applications target low-prevalence events, where apparent accuracy can conceal limited real-world value. Relevant AI models range from expert-defined rules and traditional machine learning to generative LLMs constrained for classification. We outline key considerations for critical appraisal of AI in rare-event recognition, including problem framing and test set design, prevalence-aware statistical evaluation, robustness assessment, and integration into human workflows. In addition, we propose an approach to structured case-level examination (SCLE), to complement statistical performance evaluation, and a comprehensive checklist to guide procurement or development of AI models for rare-event recognition. We instantiate the framework in pharmacovigilance, drawing on three studies: rule-based retrieval of pregnancy-related reports; duplicate detection combining machine learning with probabilistic record linkage; and automated redaction of person names using an LLM. We highlight pitfalls specific to the rare-event setting including optimism from unrealistic class balance and lack of difficult positive controls in test sets - and show how cost-sensitive targets align model performance with operational value. While grounded in pharmacovigilance practice, the principles generalize to domains where positives are scarce and error costs may be asymmetric.

Abstract--For large-scale applications, there is growing interest in replacing Graph Neural Networks (GNNs) with lightweight Multi-Layer Perceptrons (MLPs) via knowledge distillation. However, distilling GNNs for self-supervised graph representation learning into MLPs is more challenging. This is because the performance of self-supervised learning is more related to the model's inductive bias than supervised learning. This motivates us to design a new distillation method to bridge a huge capacity gap between GNNs and MLPs in self-supervised graph representation learning. In this paper, we propose Diffusion-Assisted Distillation for Self-supervised Graph representation learning with MLPs (DAD-SGM). The proposed method employs a denoising diffusion model as a teacher assistant to better distill the knowledge from the teacher GNN into the student MLP . This approach enhances the generalizability and robustness of MLPs in self-supervised graph representation learning. Extensive experiments demonstrate that DAD-SGM effectively distills the knowledge of self-supervised GNNs compared to state-of-the-art GNN-to-MLP distillation methods. Impact Statement--This paper presents Diffusion-Assisted Distillation for Self-supervised Graph representation learning with MLPs (DAD-SGM), a novel framework that addresses the performance gap between GNNs and MLPs in self-supervised graph learning. Our approach first trains an assistant denoising diffusion model that learns to predict noise from noisy outputs of the GNN teacher .

The rise of artificial intelligence (AI) as super-capable assistants has transformed productivity and decision-making across domains. Yet, this integration raises critical concerns about value alignment - ensuring AI behaviors remain consistent with human ethics and intentions. A key risk is value drift, where AI systems deviate from aligned values due to evolving contexts, learning dynamics, or unintended optimizations, potentially leading to inefficiencies or ethical breaches. We propose the Moral Anchor System (MAS), a novel framework to detect, predict, and mitigate value drift in AI agents. MAS combines real-time Bayesian inference for monitoring value states, LSTM networks for forecasting drift, and a human-centric governance layer for adaptive interventions. It emphasizes low-latency responses (<20 ms) to prevent breaches, while reducing false positives and alert fatigue via supervised fine-tuning with human feedback. Our hypothesis: integrating probabilistic drift detection, predictive analytics, and adaptive governance can reduce value drift incidents by 80 percent or more in simulations, maintaining high detection accuracy (85 percent) and low false positive rates (0.08 post-adaptation). Rigorous experiments with goal-misaligned agents validate MAS's scalability and responsiveness. MAS's originality lies in its predictive and adaptive nature, contrasting static alignment methods. Contributions include: (1) MAS architecture for AI integration; (2) empirical results prioritizing speed and usability; (3) cross-domain applicability insights; and (4) open-source code for replication.

Abstract--Class imbalance is a common challenge in real-world binary classification tasks, often leading to predictions biased toward the majority class and reduced recognition of the minority class. This issue is particularly critical in domains such as medical diagnosis and anomaly detection, where correct classification of minority classes is essential. Conventional methods often fail to deliver satisfactory performance when the imbalance ratio is extremely severe. T o address this challenge, we propose a novel approach called Kernel-density-Oriented Threshold Adjustment with Regional Optimization (KOT ARO), which extends the framework of kernel density estimation (KDE) by adaptively adjusting decision boundaries according to local sample density. In KOT ARO, the bandwidth of Gaussian basis functions is dynamically tuned based on the estimated density around each sample, thereby enhancing the classifier's ability to capture minority regions. The results demonstrated that KOT ARO outperformed conventional methods, particularly under conditions of severe imbalance, highlighting its potential as a promising solution for a wide range of imbalanced classification problems. In real-world binary classification tasks, class imbalance is a common and challenging issue, where one class significantly outnumbers the other.