This paper investigates the effectiveness of AI for fault localization using only natural-language bug reports in an industrial setting. It benchmarks traditional machine learning models against fine-tuned transformer models using proprietary data from ABB Robotics, finding that traditional models often outperform transformers in this context.

This paper introduces an autoresearch framework that uses AI coding agents to optimize ground-state preparation protocols for quantum and quantum-classical many-body systems. It demonstrates how this method can effectively tune hyperparameters for VQE, DMRG, and AFQMC protocols, leading to improved energy proxies within fixed computational budgets across various spin models and molecular Hamiltonians.

This paper introduces PREF-XAI, a novel approach for generating personalized, rule-based explanations of black-box machine learning models. It reframes explanation as a preference-driven decision problem, learning individual user preferences through robust ordinal regression to tailor explanations.

This paper introduces Probabilistic Bias Correction (PBC), a machine learning framework designed to significantly improve subseasonal weather forecasts (2-6 weeks ahead) by correcting systematic errors in existing dynamical and AI models. PBC has demonstrated superior performance in real-time forecasting competitions, enhancing predictions for temperature, pressure, and precipitation, and improving the accuracy of extreme event warnings.

This paper presents an extraction of unpolarized quark transverse-momentum-dependent parton distribution functions (TMD PDFs) from Drell-Yan data using an AI-assisted Bayesian inference framework. It performs a global analysis of Drell-Yan data and compares the results with those obtained using the replica method, highlighting differences in uncertainty estimates.

This paper reviews how the "usefulness" of representations is conceptualized across philosophy, neuroscience, and computer science. It proposes a three-level framework—Representations as Information, Usable, and Used—to organize diverse perspectives on neural representations.

This paper introduces Learning-Infused Formal Reasoning (LIFR), a framework that integrates machine learning with formal verification to address the challenges of opaque AI systems and labor-intensive formal methods. LIFR aims to enable automated contract synthesis, semantic reuse of verification artifacts, and provide rigorous semantic foundations for scalable and trustworthy software engineering.

This paper explores lightweight Generative AI (GenAI) models for network traffic synthesis to address data scarcity and privacy in Network Traffic Classification (NTC). It evaluates transformer-based, state-space, and diffusion models, demonstrating their effectiveness in generating high-fidelity synthetic traffic for training and augmenting NTC systems.

BOxCrete is an open-source AI model using Gaussian Process regression and Bayesian Optimization to predict concrete compressive strength and optimize mix designs for performance and sustainability. It addresses the complexity of modern concrete formulations by providing a reproducible, data-driven framework built on a new open-access dataset.

This paper introduces a novel PCA sweep procedure for Supervised Semantic Differential (SSD), a method modeling how text meaning varies with individual differences. The sweep systematically selects the optimal number of PCA components to ensure interpretable and stable semantic gradients, illustrated through a case study on AI discourse related to narcissism.