19332 papers
The spectrum from Cesium to Lutetium encompasses a fascinating group of elements that bridge the gap between the reactive alkali metals and the complex chemistry of the transition series. These atoms, sitting in the middle of the periodic table, drive innovations in atomic clocks, quantum computing, and advanced medical imaging. Their unique electron configurations allow scientists to probe the fundamental laws of physics with extraordinary precision, offering a window into how matter behaves under extreme conditions.
Gist.Science is dedicated to making the latest discoveries in this niche accessible to everyone. We process every new preprint in this category from arXiv as soon as it appears, translating dense academic findings into clear, plain-language summaries alongside detailed technical breakdowns. This dual approach ensures that both curious readers and specialists can grasp the significance of these rapid scientific advances without getting lost in the jargon.
Below are the latest papers covering these heavy elements and their groundbreaking applications.
This paper demonstrates that Transformer models can successfully learn and predict sequences from complex Permuted Congruential Generators (PCGs) through curriculum learning and by discovering bitwise rotationally-invariant representations, revealing a scaling law where the required context length grows as the square root of the modulus.
This paper proposes a Green AI-guided adaptive encoder freezing strategy for federated learning in MRI-to-CT conversion that significantly reduces energy consumption and CO2 emissions by up to 23% while maintaining or improving model performance, thereby promoting equitable and sustainable healthcare AI.
This paper proposes a framework and a sequential evaluation algorithm to determine the optimal timing for switching from an incumbent machine learning model to a challenger trained on new data sources, balancing the trade-off between improving predictive performance and the costs of retraining to achieve near-oracle economic efficiency.
This paper establishes the statistical foundations of amortized Bayesian inference by analyzing how major neural architectures like feedforward networks, Deep Sets, and Transformers enable efficient, low-cost posterior approximation, while empirically validating their accuracy, robustness, and uncertainty quantification across diverse simulation scenarios.
This paper introduces Spectron, a spectral renormalization technique that stabilizes end-to-end native low-rank LLM pretraining by controlling weight update growth, enabling models trained from scratch with exclusively low-rank weights to match dense model performance while improving inference efficiency.
This paper introduces ARTIST, a reinforcement learning-based framework that treats time series reasoning as a sequential decision problem to dynamically select informative temporal segments, thereby significantly outperforming existing models by avoiding static full-sequence encoding and focusing on task-relevant information.
This study presents a YOLO-based deep learning framework integrated with High-Resolution Class Activation Mapping (HiResCAM) to achieve over 96% accuracy in the automated, interpretable identification of Ichneumonoidea wasp families from high-resolution images, thereby addressing the challenges of manual taxonomic identification in biodiversity and biological control programs.
This paper introduces Adversarial Dynamics Priors (ADP), a method that replaces kinematic motion features with physically grounded dynamics features in adversarial regularization to significantly enhance the perturbation resilience and recovery performance of humanoid locomotion control compared to existing motion-tracking approaches.
This paper presents an application-oriented framework for generative AI in finance, organizing its capabilities into five key patterns mapped to major financial functions and technical architectures to identify high-impact use cases and establish a foundation for effective risk control.
This position paper argues that the Explainable AI community must shift its focus from developing ad-hoc techniques to addressing foundational structural challenges—such as unclear problem formulations and the lack of integration pipelines—to enable explanations to effectively drive real-world, human-in-the-loop actions.