At first glance at this creature, you'd be forgiven for mistaking it for a sparkly pair of fake eyelashes. But the creature is very much real and was discovered at the junction of three tectonic plates in the Pacific Ocean. Researchers from the Schmidt Ocean Institute spotted the animal while using an underwater robot to scour the seabed. The animal is a polychaete - a class of marine worms, more widely known as bristle worms. 'To describe this polychaete, one simply must use jazz hands -- it is the only way to capture this deep-sea worm's dazzle!' the experts said in an Instagram post about the polychaete.

Throughout the scientific computing space, deep learning algorithms have shown excellent performance in a wide range of applications. As these deep neural networks (DNNs) continue to mature, the necessary compute required to train them has continued to grow. Today, modern DNNs require millions of FLOPs and days to weeks of training to generate a well-trained model. The training times required for DNNs are oftentimes a bottleneck in DNN research for a variety of deep learning applications, and as such, accelerating and scaling DNN training enables more robust and accelerated research. To that end, in this work, we explore utilizing the NRP Nautilus HyperCluster to automate and scale deep learning model training for three separate applications of DNNs, including overhead object detection, burned area segmentation, and deforestation detection. In total, 234 deep neural models are trained on Nautilus, for a total time of 4,040 hours. Deep convolutional neural networks (DCNNs) have been established as the state of the art in computer vision (CV) and have shown superior performance in visual tasks for many domains, including remote sensing. With billions of pixels being collected by overhead sources like satellites, remote sensing (RS) is becoming evermore a big-data problem domain, with endless amounts of data available to enable CV applications. Due in part to this data availability, the training and optimization of deep networks for RS applications has been explored to great lengths in recent years. In 2017, researchers investigated utilizing DCNNs for land-cover classification in overhead imagery along with techniques such as transfer learning and data augmentation[1]. This work was then extended into multi-network fusion research, where multiple DCNNs trained on overhead satellite imagery were fused using simple fusion techniques such as voting and arrogance [2] and then compared to more complex fusion algorithms such as the Choquet and Sugeno Fuzzy Integral [3], [4]. While these studies explored utilizing DCNNs to perform classification on overhead RS imagery, further exploration was required in broad area search, in which DCNNs are trained and used not on clean pre-processed datasets, but instead applied to large swaths of overhead imagery with the goal of finding all instances of a given object or terrain.

This paper builds upon our previous work on the Reconciled Polynomial Network (RPN). The original RPN model was designed under the assumption of input data independence, presuming the independence among both individual instances within data batches and attributes in each data instance. However, this assumption often proves invalid for function learning tasks involving complex, interdependent data such as language, images, time series, and graphs. Ignoring such data interdependence may inevitably lead to significant performance degradation. To overcome these limitations, we introduce the new Reconciled Polynomial Network (version 2), namely RPN 2, in this paper. By incorporating data and structural interdependence functions, RPN 2 explicitly models data interdependence via new component functions in its architecture. This enhancement not only significantly improves RPN 2's learning performance but also substantially expands its unifying potential, enabling it to encompass a broader range of contemporary dominant backbone models within its canonical representation. These backbones include, but are not limited to, convolutional neural networks (CNNs), recurrent neural networks (RNNs), graph neural networks (GNNs), and Transformers. Our analysis reveals that the fundamental distinctions among these backbone models primarily stem from their diverse approaches to defining the interdependence functions. Furthermore, this unified representation opens up new opportunities for designing innovative architectures with the potential to surpass the performance of these dominant backbones.

This paper introduces the \emph{Optimist}, an autonomous system developed to advance automated conjecture generation in graph theory. Leveraging mixed-integer programming (MIP) and heuristic methods, the \emph{Optimist} generates conjectures that both rediscover established theorems and propose novel inequalities. Through a combination of memory-based computation and agent-like adaptability, the \emph{Optimist} iteratively refines its conjectures by integrating new data, enabling a feedback process with minimal human (\emph{or machine}) intervention. Initial experiments reveal the \emph{Optimist}'s potential to uncover foundational results in graph theory, as well as to produce conjectures of interest for future exploration. This work also outlines the \emph{Optimist}'s evolving integration with a counterpart agent, the \emph{Pessimist} (a human \emph{or machine} agent), to establish a dueling system that will drive fully automated graph theory research.

The remarkable progress in neural-network-driven visual data generation, especially with neural rendering techniques like Neural Radiance Fields and 3D Gaussian splatting, offers a powerful alternative to GANs and diffusion models. These methods can produce high-fidelity images and lifelike avatars, highlighting the need for robust detection methods. In response, an unsupervised training technique is proposed that enables the model to extract comprehensive features from the Fourier spectrum magnitude, thereby overcoming the challenges of reconstructing the spectrum due to its centrosymmetric properties. By leveraging the spectral domain and dynamically combining it with spatial domain information, we create a robust multimodal detector that demonstrates superior generalization capabilities in identifying challenging synthetic images generated by the latest image synthesis techniques. To address the absence of a 3D neural rendering-based fake image database, we develop a comprehensive database that includes images generated by diverse neural rendering techniques, providing a robust foundation for evaluating and advancing detection methods.

Retrieval-augmented generation (RAG) is a central component of modern LLM systems, particularly in scenarios where up-to-date information is crucial for accurately responding to user queries or when queries exceed the scope of the training data. The advent of time-series foundation models (TSFM), such as Chronos, and the need for effective zero-shot forecasting performance across various time-series domains motivates the question: Do benefits of RAG similarly carry over to time series forecasting? In this paper, we advocate that the dynamic and event-driven nature of time-series data makes RAG a crucial component of TSFMs and introduce a principled RAG framework for time-series forecasting, called Retrieval Augmented Forecasting (RAF). Within RAF, we develop efficient strategies for retrieving related time-series examples and incorporating them into forecast. Through experiments and mechanistic studies, we demonstrate that RAF indeed improves the forecasting accuracy across diverse time series domains and the improvement is more significant for larger TSFM sizes.

As urbanization and climate change progress, urban heat becomes a priority for climate adaptation efforts. High temperatures concentrated in urban heat can drive increased risk of heat-related death and illness as well as increased energy demand for cooling. However, estimating the effects of urban heat is an ongoing field of research typically burdened by an imprecise description of the built environment, significant computational cost, and a lack of high-resolution estimates of the impacts of climate change. Here, we present open-source, computationally efficient machine learning methods that can improve the accuracy of urban temperature estimates when compared to historical reanalysis data. These models are applied to residential buildings in Los Angeles, and we compare the energy benefits of heat mitigation strategies to the impacts of climate change. We find that cooling demand is likely to increase substantially through midcentury, but engineered high-albedo surfaces could lessen this increase by more than 50%. The corresponding increase in heating demand complicates this narrative, but total annual energy use from combined heating and cooling with electric heat pumps in the Los Angeles urban climate is shown to benefit from the engineered cooling strategies under both current and future climates.

The exponential growth of multivariate time series data from sensor networks in domains like industrial monitoring and smart cities requires efficient and accurate forecasting models. Current deep learning methods often fail to adequately capture long-range dependencies and complex inter-variable relationships, especially under real-time processing constraints. These limitations arise as many models are optimized for either short-term forecasting with limited receptive fields or long-term accuracy at the cost of efficiency. Additionally, dynamic and intricate interactions between variables in real-world data further complicate modeling efforts. To address these limitations, we propose EffiCANet, an Efficient Convolutional Attention Network designed to enhance forecasting accuracy while maintaining computational efficiency. EffiCANet integrates three key components: (1) a Temporal Large-kernel Decomposed Convolution (TLDC) module that captures long-term temporal dependencies while reducing computational overhead; (2) an Inter-Variable Group Convolution (IVGC) module that captures complex and evolving relationships among variables; and (3) a Global Temporal-Variable Attention (GTVA) mechanism that prioritizes critical temporal and inter-variable features. Extensive evaluations across nine benchmark datasets show that EffiCANet achieves the maximum reduction of 10.02% in MAE over state-of-the-art models, while cutting computational costs by 26.2% relative to conventional large-kernel convolution methods, thanks to its efficient decomposition strategy.

We sketch how developers of frontier AI systems could construct a structured rationale -- a 'safety case' -- that an AI system is unlikely to cause catastrophic outcomes through scheming. Scheming is a potential threat model where AI systems could pursue misaligned goals covertly, hiding their true capabilities and objectives. In this report, we propose three arguments that safety cases could use in relation to scheming. For each argument we sketch how evidence could be gathered from empirical evaluations, and what assumptions would need to be met to provide strong assurance. First, developers of frontier AI systems could argue that AI systems are not capable of scheming (Scheming Inability). Second, one could argue that AI systems are not capable of posing harm through scheming (Harm Inability). Third, one could argue that control measures around the AI systems would prevent unacceptable outcomes even if the AI systems intentionally attempted to subvert them (Harm Control). Additionally, we discuss how safety cases might be supported by evidence that an AI system is reasonably aligned with its developers (Alignment). Finally, we point out that many of the assumptions required to make these safety arguments have not been confidently satisfied to date and require making progress on multiple open research problems.

With the growing complexity and capability of large language models, a need to understand model reasoning has emerged, often motivated by an underlying goal of controlling and aligning models. While numerous interpretability and steering methods have been proposed as solutions, they are typically designed either for understanding or for control, seldom addressing both, with the connection between interpretation and control more broadly remaining tenuous. Additionally, the lack of standardized applications, motivations, and evaluation metrics makes it difficult to assess these methods' practical utility and efficacy. To address this, we propose intervention as a fundamental goal of interpretability and introduce success criteria to evaluate how well methods are able to control model behavior through interventions. We unify and extend four popular interpretability methods--sparse autoencoders, logit lens, tuned lens, and probing--into an abstract encoder-decoder framework. This framework maps intermediate latent representations to human-interpretable feature spaces, enabling interventions on these interpretable features, which can then be mapped back to latent representations to control model outputs. We introduce two new evaluation metrics: intervention success rate and the coherence-intervention tradeoff, designed to measure the accuracy of explanations and their utility in controlling model behavior. Our findings reveal that (1) although current methods allow for intervention, they are inconsistent across models and features, (2) lens-based methods outperform others in achieving simple, concrete interventions, and (3) interventions often compromise model performance and coherence, underperforming simpler alternatives, such as prompting, for steering model behavior and highlighting a critical shortcoming of current interpretability approaches in real-world applications requiring control.