Abstract
Imagine decision-makers uploading data and, within minutes, receiving clear, actionable insights delivered straight to their fingertips. That is the promise of the AI Data Scientist, an autonomous Agent powered by large language models (LLMs) that closes the gap between evidence and action. Rather than simply writing code or responding to prompts, it reasons through questions, tests ideas, and delivers end-to-end insights at a pace far beyond traditional workflows. Guided by the scientific tenet of the hypothesis, this Agent uncovers explanatory patterns in data, evaluates their statistical significance, and uses them to inform predictive modeling. It then translates these results into recommendations that are both rigorous and accessible. At the core of the AI Data Scientist is a team of specialized LLM Subagents, each responsible for a distinct task such as data cleaning, statistical testing, validation, and plain-language communication. These Subagents write their own code, reason about causality, and identify when additional data is needed to support sound conclusions. Together, they achieve in minutes what might otherwise take days or weeks, enabling a new kind of interaction that makes deep data science both accessible and actionable.

Abstract
Competitions are widely used to identify top performers in judgmental forecasting and machine learning, and the standard competition design ranks competitors based on their cumulative scores against a set of realized outcomes or held-out labels. However, this standard design is neither incentive-compatible nor very statistically efficient. The main culprit is noise in outcomes/labels that experts are scored against; it allows weaker competitors to often win by chance, and the winner-take-all nature incentivizes misreporting that improves win probability even if it decreases expected score. Attempts to achieve incentive-compatibility rely on randomized mechanisms that add even more noise in winner selection, but come at the cost of determinism and practical adoption. To tackle these issues, we introduce a novel deterministic mechanism: WOMAC (Wisdom of the Most Accurate Crowd). Instead of scoring experts against noisy outcomes, as is standard, WOMAC scores experts against the best ex-post aggregate of peer experts' predictions given the noisy outcomes. WOMAC is also more efficient than the standard competition design in typical settings. While the increased complexity of WOMAC makes it challenging to analyze incentives directly, we provide a clear theoretical foundation to justify the mechanism. We also provide an efficient vectorized implementation and demonstrate empirically on real-world forecasting datasets that WOMAC is a more reliable predictor of experts' out-of-sample performance relative to the standard mechanism. WOMAC is useful in any competition where there is substantial noise in the outcomes/labels.

Abstract
Unit testing is a core practice in programming, enabling systematic evaluation of programs produced by human developers or large language models (LLMs). Given the challenges in writing comprehensive unit tests, LLMs have been employed to automate test generation, yet methods for training LLMs to produce high-quality tests remain underexplored. In this work, we propose UTRL, a novel reinforcement learning framework that trains an LLM to generate high-quality unit tests given a programming instruction. Our key idea is to iteratively train two LLMs, the unit test generator and the code generator, in an adversarial manner via reinforcement learning. The unit test generator is trained to maximize a discrimination reward, which reflects its ability to produce tests that expose faults in the code generator's solutions, and the code generator is trained to maximize a code reward, which reflects its ability to produce solutions that pass the unit tests generated by the test generator. In our experiments, we demonstrate that unit tests generated by Qwen3-4B trained via UTRL show higher quality compared to unit tests generated by the same model trained via supervised fine-tuning on human-written ground-truth unit tests, yielding code evaluations that more closely align with those induced by the ground-truth tests. Moreover, Qwen3-4B trained with UTRL outperforms frontier models such as GPT-4.1 in generating high-quality unit tests, highlighting the effectiveness of UTRL in training LLMs for this task.

Abstract
We present a first known prototype of a dynamic, systematic benchmark of medical guidelines for 400+ questions, with 3.3+ trillion possible combinations, covering 100\% of guideline relationships. We transformed the WHO IMCI handbook into a directed graph with 200+ nodes (conditions, symptoms, treatments, follow-ups, severities) and 300+ edges, then used graph traversal to generate questions that incorporated age-specific scenarios and contextual distractors to ensure clinical relevance. Our graph-based approach enables systematic evaluation across clinical tasks (45-67\% accuracy), and we find models excel at symptom recognition but struggle with triaging severity, treatment protocols and follow-up care, demonstrating how customized benchmarks can identify specific capability gaps that general-domain evaluations miss. Beyond evaluation, this dynamic MCQA methodology enhances LLM post-training (supervised finetuning, GRPO, DPO), where correct answers provide high-reward samples without expensive human annotation. The graph-based approach successfully addresses the coverage limitations of manually curated benchmarks. This methodology is a step toward scalable, contamination-resistant solution for creating comprehensive benchmarks that can be dynamically generated, including when the guidelines are updated. Code and datasets are available at https://github.com/jessicalundin/graph_testing_harness

Abstract
Data-driven decisions shape public health policies and practice, yet persistent disparities in data representation skew insights and undermine interventions. To address this, we advance a structured roadmap that integrates public health data science with computer science and is grounded in reflexivity. We adopt data equity as a guiding concept: ensuring the fair and inclusive representation, collection, and use of data to prevent the introduction or exacerbation of systemic biases that could lead to invalid downstream inference and decisions. To underscore urgency, we present three public health cases where non-representative datasets and skewed knowledge impede decisions across diverse subgroups. These challenges echo themes in two literatures: public health highlights gaps in high-quality data for specific populations, while computer science and statistics contribute criteria and metrics for diagnosing bias in data and models. Building on these foundations, we propose a working definition of public health data equity and a structured self-audit framework. Our framework integrates core computational principles (fairness, accountability, transparency, ethics, privacy, confidentiality) with key public health considerations (selection bias, representativeness, generalizability, causality, information bias) to guide equitable practice across the data life cycle, from study design and data collection to measurement, analysis, interpretation, and translation. Embedding data equity in routine practice offers a practical path for ensuring that data-driven policies, artificial intelligence, and emerging technologies improve health outcomes for all. Finally, we emphasize the critical understanding that, although data equity is an essential first step, it does not inherently guarantee information, learning, or decision equity.

Abstract
We consider the problem of sorting $n$ elements subject to persistent random comparison errors. In this problem, each comparison between two elements can be wrong with some fixed (small) probability $p$, and comparing the same pair of elements multiple times always yields the same result. Sorting perfectly in this model is impossible, and the objective is to minimize the dislocation of each element in the output sequence, i.e., the difference between its position in the sequence and its true rank. In this paper, we present the first $O(n\log n)$-time sorting algorithm that guarantees both $O(\log n)$ maximum dislocation and $O(n)$ total dislocation with high probability when $p<\frac{1}{4}$. This settles the time complexity sorting with persistent comparison errors in the given range of $p$ and shows that comparison errors do not increase its computational difficulty. Indeed, $\Omega(n\log n)$ time is necessary to archive a maximum dislocation of $O(\log n)$ even without comparison errors. Moreover, we prove that no algorithm can guarantee a maximum dislocation of $o(\log n)$ with high probability, nor a total dislocation of $o(n)$ in expectation. To develop our sorting algorithm, we solve two related sub-problems, which might be of independent interest. More precisely, we show that $O(\log n)$ time suffices to find a position in which to insert a new element $x$ in an almost-sorted sequence $S$ of $n$ elements having dislocation at most $d=\Omega(\log n)$, so that the dislocation of $x$ in the resulting sequence is $O(d)$ with high probability (which can be equivalently thought as the problem of estimating the rank of $x$ in $S$). We also show that the maximum (resp. total) dislocation of an approximately sorted sequence $S$ of $n$ elements can be lowered to $O(\log n)$ (resp. $O(n)$) in $O(nd)$ time, w.h.p., where $d$ is an upper bound on the maximum dislocation of $S$.

Abstract
In safety-critical deep learning applications, robustness measures the ability of neural models that handle imperceptible perturbations in input data, which may lead to potential safety hazards. Existing pre-deployment robustness assessment methods typically suffer from significant trade-offs between computational cost and measurement precision, limiting their practical utility. To address these limitations, this paper conducts a comprehensive comparative analysis of existing robustness definitions and associated assessment methodologies. We propose tower robustness to evaluate robustness, which is a novel, practical metric based on hypothesis testing to quantitatively evaluate probabilistic robustness, enabling more rigorous and efficient pre-deployment assessments. Our extensive comparative evaluation illustrates the advantages and applicability of our proposed approach, thereby advancing the systematic understanding and enhancement of model robustness in safety-critical deep learning applications.

Abstract
Machine learning (ML) is increasingly applied across industries to automate decision-making, but concerns about ethical and legal compliance remain due to limited transparency, fairness, and accountability. Monitoring through logging a long-standing practice in traditional software offers a potential means for auditing ML applications, as logs provide traceable records of system behavior useful for debugging, performance analysis, and continuous auditing. systematically auditing models for compliance or accountability. The findings underscore the need for enhanced logging practices and tooling that systematically integrate responsible AI metrics. Such practices would support the development of auditable, transparent, and ethically responsible ML systems, aligning with growing regulatory requirements and societal expectations. By highlighting specific deficiencies and opportunities, this work provides actionable guidance for both practitioners and tool developers seeking to strengthen the accountability and trustworthiness of ML applications.

Abstract
This study explores the deployment of three machine learning (ML) approaches for real-time prediction of glucose, lactate, and ammonium concentrations in cell culture processes, using Raman spectroscopy as input features. The research addresses challenges associated with limited data availability and process variability, providing a comparative analysis of pretrained models, just-in-time learning (JITL), and online learning algorithms. Two industrial case studies are presented to evaluate the impact of varying bioprocess conditions on model performance. The findings highlight the specific conditions under which pretrained models demonstrate superior predictive accuracy and identify scenarios where JITL or online learning approaches are more effective for adaptive process monitoring. This study also highlights the critical importance of updating the deployed models/agents with the latest offline analytical measurements during bioreactor operations to maintain the model performance against the changes in cell growth behaviours and operating conditions throughout the bioreactor run. Additionally, the study confirms the usefulness of a simple mixture-of-experts framework in achieving enhanced accuracy and robustness for real-time predictions of metabolite concentrations based on Raman spectral data. These insights contribute to the development of robust strategies for the efficient deployment of ML models in dynamic and changing biomanufacturing environments.

Abstract
As artificial intelligence (AI) becomes foundational to enterprise infrastructure, organizations face growing challenges in accurately assessing the full economic implications of AI deployment. Existing metrics such as API token costs, GPU-hour billing, or Total Cost of Ownership (TCO) fail to capture the complete lifecycle costs of AI systems and provide limited comparability across deployment models. This paper introduces the Levelized Cost of Artificial Intelligence (LCOAI), a standardized economic metric designed to quantify the total capital (CAPEX) and operational (OPEX) expenditures per unit of productive AI output, normalized by valid inference volume. Analogous to established metrics like LCOE (levelized cost of electricity) and LCOH (levelized cost of hydrogen) in the energy sector, LCOAI offers a rigorous, transparent framework to evaluate and compare the cost-efficiency of vendor API deployments versus self-hosted, fine-tuned models. We define the LCOAI methodology in detail and apply it to three representative scenarios, OpenAI GPT-4.1 API, Anthropic Claude Haiku API, and a self-hosted LLaMA-2-13B deployment demonstrating how LCOAI captures critical trade-offs in scalability, investment planning, and cost optimization. Extensive sensitivity analyses further explore the impact of inference volume, CAPEX, and OPEX variability on lifecycle economics. The results illustrate the practical utility of LCOAI in procurement, infrastructure planning, and automation strategy, and establish it as a foundational benchmark for AI economic analysis. Policy implications and areas for future refinement, including environmental and performance-adjusted cost metrics, are also discussed.

Abstract
Training large language models faces frequent interruptions due to various faults, demanding robust fault-tolerance. Existing backup-free methods, such as redundant computation, dynamic parallelism, and data rerouting, each incur performance penalties, whether from ongoing overhead, lengthy reconfigurations, or post-recovery inefficiencies. We propose Odyssey, an adaptive fault-tolerant system that intelligently selects optimal recovery strategies when a failure occurs. Odyssey achieves this through a unified performance model, expedient execution plan search, accurate performance estimation, and efficient communication optimizations. Experiments on a 32-card cluster show that Odyssey maintains a performance gap of within 11.00% between post-recovery and failure-free training, while preserving model convergence and efficient memory usage. Compared to state-of-the-art methods, Odyssey achieves up to 1.229x and 1.355x higher average throughput than Oobleck and Recycle, respectively.

Abstract
Minimizing computational overhead in time-series classification, particularly in deep learning models, presents a significant challenge due to the high complexity of model architectures and the large volume of sequential data that must be processed in real time. This challenge is further compounded by adversarial attacks, emphasizing the need for resilient methods that ensure robust performance and efficient model selection. To address this challenge, we propose ReLATE+, a comprehensive framework that detects and classifies adversarial attacks, adaptively selects deep learning models based on dataset-level similarity, and thus substantially reduces retraining costs relative to conventional methods that do not leverage prior knowledge, while maintaining strong performance. ReLATE+ first checks whether the incoming data is adversarial and, if so, classifies the attack type, using this insight to identify a similar dataset from a repository and enable the reuse of the best-performing associated model. This approach ensures strong performance while reducing the need for retraining, and it generalizes well across different domains with varying data distributions and feature spaces. Experiments show that ReLATE+ reduces computational overhead by an average of 77.68%, enhancing adversarial resilience and streamlining robust model selection, all without sacrificing performance, within 2.02% of Oracle.

Abstract
The demand for text classification is growing significantly in web searching, data mining, web ranking, recommendation systems, and so many other fields of information and technology. This paper illustrates the text classification process on different datasets using some standard supervised machine learning techniques. Text documents can be classified through various kinds of classifiers. Labeled text documents are used to classify the text in supervised classifications. This paper applies these classifiers on different kinds of labeled documents and measures the accuracy of the classifiers. An Artificial Neural Network (ANN) model using Back Propagation Network (BPN) is used with several other models to create an independent platform for labeled and supervised text classification process. An existing benchmark approach is used to analyze the performance of classification using labeled documents. Experimental analysis on real data reveals which model works well in terms of classification accuracy.

Abstract
Machine learning algorithms can now outperform classic economic models in predicting quantities ranging from bargaining outcomes, to choice under uncertainty, to an individual's future jobs and wages. Yet this predictive accuracy comes at a cost: most machine learning algorithms function as black boxes, offering little insight into \emph{why} outcomes occur. This article asks whether machine learning can guide the development of new economic theories. Economic models serve an important purpose beyond prediction -- they uncover the general mechanisms behind observed behaviors. A model that identifies the causal pathways of economic development is more valuable than one that merely predicts which countries will escape poverty, because it enables policymakers to encourage that development in countries where it might not have happened otherwise. Similarly, a model that predicts imperfectly across many domains can be more valuable than one that is highly accurate in a specific domain, since the former allows insights and data obtained from one setting to inform decisions and policy in another. Applying machine learning algorithms off-the-shelf is unlikely to yield such models. But recent work shows that, when reconceived with the aims of an economic modeler in mind, machine learning methods can improve both prediction and understanding. These approaches range from adversarially training algorithms to expose the limits of existing models, to imposing economic theory as a constraint on algorithmic search. Advances in large language models complement these strategies and open new research directions.

Abstract
The growing adoption of foundation models calls for a paradigm shift from Data Science to Model Science. Unlike data-centric approaches, Model Science places the trained model at the core of analysis, aiming to interact, verify, explain, and control its behavior across diverse operational contexts. This paper introduces a conceptual framework for a new discipline called Model Science, along with the proposal for its four key pillars: Verification, which requires strict, context-aware evaluation protocols; Explanation, which is understood as various approaches to explore of internal model operations; Control, which integrates alignment techniques to steer model behavior; and Interface, which develops interactive and visual explanation tools to improve human calibration and decision-making. The proposed framework aims to guide the development of credible, safe, and human-aligned AI systems.

Abstract
Model updates (new hyperparameters, kernels, depths, solvers, or data) change performance, but the \emph{reason} often remains opaque. We introduce \textbf{Delta-Attribution} (\mbox{$\Delta$-Attribution}), a model-agnostic framework that explains \emph{what changed} between versions $A$ and $B$ by differencing per-feature attributions: $\Delta\phi(x)=\phi_B(x)-\phi_A(x)$. We evaluate $\Delta\phi$ with a \emph{$\Delta$-Attribution Quality Suite} covering magnitude/sparsity (L1, Top-$k$, entropy), agreement/shift (rank-overlap@10, Jensen--Shannon divergence), behavioural alignment (Delta Conservation Error, DCE; Behaviour--Attribution Coupling, BAC; CO$\Delta$F), and robustness (noise, baseline sensitivity, grouped occlusion). Instantiated via fast occlusion/clamping in standardized space with a class-anchored margin and baseline averaging, we audit 45 settings: five classical families (Logistic Regression, SVC, Random Forests, Gradient Boosting, $k$NN), three datasets (Breast Cancer, Wine, Digits), and three A/B pairs per family. \textbf{Findings.} Inductive-bias changes yield large, behaviour-aligned deltas (e.g., SVC poly$\!\rightarrow$rbf on Breast Cancer: BAC$\approx$0.998, DCE$\approx$6.6; Random Forest feature-rule swap on Digits: BAC$\approx$0.997, DCE$\approx$7.5), while ``cosmetic'' tweaks (SVC \texttt{gamma=scale} vs.\ \texttt{auto}, $k$NN search) show rank-overlap@10$=1.0$ and DCE$\approx$0. The largest redistribution appears for deeper GB on Breast Cancer (JSD$\approx$0.357). $\Delta$-Attribution offers a lightweight update audit that complements accuracy by distinguishing benign changes from behaviourally meaningful or risky reliance shifts.

Abstract
AI researchers and practitioners increasingly apply large language models (LLMs) to what we call reasoning-intensive regression (RiR), i.e. deducing subtle numerical properties from text. Unlike standard language regression tasks, e.g. for sentiment or similarity, RiR often appears instead in ad-hoc problems like rubric-based scoring or domain-specific retrieval, where much deeper analysis of text is required while only limited task-specific training data and computation are available. We cast three realistic problems as RiR tasks to establish an initial benchmark, and use that to test our hypothesis that prompting frozen LLMs and finetuning Transformer encoders via gradient descent will both often struggle in RiR. We then propose MENTAT, a simple and lightweight method that combines batch-reflective prompt optimization with neural ensemble learning. MENTAT achieves up to 65% improvement over both baselines, though substantial room remains for future advances in RiR.

Abstract
Current inference systems for Mixture-of-Experts (MoE) models primarily employ static parallelization strategies. However, these static approaches cannot consistently achieve optimal performance across different inference scenarios, as they lack the flexibility to adapt to varying computational requirements. In this work, we propose HAP (Hybrid Adaptive Parallelism), a novel method that dynamically selects hybrid parallel strategies to enhance MoE inference efficiency. The fundamental innovation of HAP lies in hierarchically decomposing MoE architectures into two distinct computational modules: the Attention module and the Expert module, each augmented with a specialized inference latency simulation model. This decomposition promotes the construction of a comprehensive search space for seeking model parallel strategies. By leveraging Integer Linear Programming (ILP), HAP could solve the optimal hybrid parallel configurations to maximize inference efficiency under varying computational constraints. Our experiments demonstrate that HAP consistently determines parallel configurations that achieve comparable or superior performance to the TP strategy prevalent in mainstream inference systems. Compared to the TP-based inference, HAP-based inference achieves speedups of 1.68x, 1.77x, and 1.57x on A100, A6000, and V100 GPU platforms, respectively. Furthermore, HAP showcases remarkable generalization capability, maintaining performance effectiveness across diverse MoE model configurations, including Mixtral and Qwen series models.