We present the first language-model-driven agentic artificial intelligence (AI) system to autonomously execute multi-stage physics experiments on a production synchrotron light source. Implemented at the Advanced Light Source particle accelerator, the system translates natural language user prompts into structured execution plans that combine archive data retrieval, control-system channel resolution, automated script generation, controlled machine interaction, and analysis. In a representative machine physics task, we show that preparation time was reduced by two orders of magnitude relative to manual scripting even for a system expert, while operator-standard safety constraints were strictly upheld. Core architectural features, plan-first orchestration, bounded tool access, and dynamic capability selection, enable transparent, auditable execution with fully reproducible artifacts. These results establish a blueprint for the safe integration of agentic AI into accelerator experiments and demanding machine physics studies, as well as routine operations, with direct portability across accelerators worldwide and, more broadly, to other large-scale scientific infrastructures.

Abstract--Pre-training strategies play a critical role in advancing the performance of transformer-based models for 3D point cloud tasks. In this paper, we introduce Point-RTD (Replaced T oken Denoising), a novel pretraining strategy designed to improve token robustness through a corruption-reconstruction framework. Unlike traditional mask-based reconstruction tasks that hide data segments for later prediction, Point-RTD corrupts point cloud tokens and leverages a discriminator-generator architecture for denoising. This shift enables more effective learning of structural priors and significantly enhances model performance and efficiency. On the ShapeNet dataset, Point-RTD reduces reconstruction error by over 93% compared to PointMAE, and achieves more than 14 lower Chamfer Distance on the test set. Point clouds have become an essential data representation in various fields such as remote sensing, autonomous driving, and robotics [1].

Time series data is a prevalent form of data found in various fields. It consists of a series of measurements taken over time. Forecasting is a crucial application of time series models, where future values are predicted based on historical data. Accurate forecasting is essential for making well-informed decisions across industries. When it comes to electric vehicles (EVs), precise predictions play a key role in planning infrastructure development, load balancing, and energy management. This study introduces a BI-LSTM embedding denoising autoencoder model (BDM) designed to address time series problems, focusing on short-term EV charging load prediction. The performance of the proposed model is evaluated by comparing it with benchmark models like Transformer, CNN, RNN, LSTM, and GRU. Based on the results of the study, the proposed model outperforms the benchmark models in four of the five-time steps, demonstrating its effectiveness for time series forecasting. This research makes a significant contribution to enhancing time series forecasting, thereby improving decision-making processes.

To address the intermittency of renewable energy source (RES) generation, scenario forecasting offers a series of stochastic realizations for predictive objects with superior flexibility and direct views. Based on a long time-series perspective, this paper explores uncertainties in the realms of renewable power and deep learning. Then, an uncertainty-aware model is meticulously designed for renewable scenario forecasting, which leverages an attention mechanism and generative adversarial networks (GANs) to precisely capture complex spatial-temporal dynamics. To improve the interpretability of uncertain behavior in RES generation, Bayesian deep learning and adaptive instance normalization (AdaIN) are incorporated to simulate typical patterns and variations. Additionally, the integration of meteorological information, forecasts, and historical trajectories in the processing layer improves the synergistic forecasting capability for multiscale periodic regularities. Numerical experiments and case analyses demonstrate that the proposed approach provides an appropriate interpretation for renewable uncertainty representation, including both aleatoric and epistemic uncertainties, and shows superior performance over state-of-the-art methods.

The growing demand for reliable electricity in universities necessitates intelligent energy management. This study proposes a machine learning-based load shedding framework for the University of Lagos, designed to optimize distribution and reduce waste. The methodology followed three main stages. First, a dataset of 3,648 hourly records from 55 buildings was compiled to develop building-level consumption models. Second, Principal Component Analysis was applied for dimensionality reduction, and clustering validation techniques were used to determine the optimal number of demand groups. Mini-Batch K-Means was then employed to classify buildings into high-, medium-, and low-demand clusters. Finally, short-term load forecasting was performed at the cluster level using multiple statistical and deep learning models, including ARIMA, SARIMA, Prophet, LSTM, and GRU. Results showed Prophet offered the most reliable forecasts, while Mini-Batch K-Means achieved stable clustering performance. By integrating clustering with forecasting, the framework enabled a fairer, data-driven load shedding strategy that reduces inefficiencies and supports climate change mitigation through sustainable energy management.

The integration of solar power has been increasing as the green energy transition rolls out. The penetration of solar power challenges the grid stability and energy scheduling, due to its intermittent energy generation. Accurate and near real-time solar power prediction is of critical importance to tolerant and support the permeation of distributed and volatile solar power production in the energy system. In this paper, we propose a deep-learning based ultra-short-term solar power prediction with data reconstruction. We decompose the data for the prediction to facilitate extensive exploration of the spatial and temporal dependencies within the data. Particularly, we reconstruct the data into low- and high-frequency components, using ensemble empirical model decomposition with adaptive noise (CEEMDAN). We integrate meteorological data with those two components, and employ deep-learning models to capture long- and short-term dependencies towards the target prediction period. In this way, we excessively exploit the features in historical data in predicting a ultra-short-term solar power production. Furthermore, as ultra-short-term prediction is vulnerable to local optima, we modify the optimization in our deep-learning training by penalizing long prediction intervals. Numerical experiments with diverse settings demonstrate that, compared to baseline models, the proposed method achieves improved generalization in data reconstruction and higher prediction accuracy for ultra-short-term solar power production.

Recently, deep learning has driven significant advancements in multivariate time series forecasting (MTSF) tasks. However, much of the current research in MTSF tends to evaluate models from a holistic perspective, which obscures the individual contributions and leaves critical issues unaddressed. Adhering to the current modeling paradigms, this work bridges these gaps by systematically decomposing deep MTSF methods into their core, fine-grained components like series-patching tokenization, channel-independent strategy, attention modules, or even Large Language Models and Time-series Foundation Models. Through extensive experiments and component-level analysis, our work offers more profound insights than previous benchmarks that typically discuss models as a whole. Furthermore, we propose a novel automated solution called TSGym for MTSF tasks. Unlike traditional hyperparameter tuning, neural architecture searching or fixed model selection, TSGym performs fine-grained component selection and automated model construction, which enables the creation of more effective solutions tailored to diverse time series data, therefore enhancing model transferability across different data sources and robustness against distribution shifts. Extensive experiments indicate that TSGym significantly outperforms existing state-of-the-art MTSF and AutoML methods. All code is publicly available on https://github.com/SUFE-AILAB/TSGym.

Recent studies have suggested a processing framework for multilingual inputs in decoder-based LLMs: early layers convert inputs into English-centric and language-agnostic representations; middle layers perform reasoning within an English-centric latent space; and final layers generate outputs by transforming these representations back into language-specific latent spaces. However, the internal dynamics of such transformation and the underlying mechanism remain underexplored. Towards a deeper understanding of this framework, we propose and empirically validate The Transfer Neurons Hypothesis: certain neurons in the MLP module are responsible for transferring representations between language-specific latent spaces and a shared semantic latent space. Furthermore, we show that one function of language-specific neurons, as identified in recent studies, is to facilitate movement between latent spaces. Finally, we show that transfer neurons are critical for reasoning in multilingual LLMs.

This paper presents a novel Koopman operator formulation for Euler Lagrangian dynamics that employs an implicit generalized momentum-based state space representation, which decouples a known linear actuation channel from state dependent dynamics and makes the system more amenable to linear Koopman modeling. By leveraging this structural separation, the proposed formulation only requires to learn the unactuated dynamics rather than the complete actuation dependent system, thereby significantly reducing the number of learnable parameters, improving data efficiency, and lowering overall model complexity. In contrast, conventional explicit formulations inherently couple inputs with the state dependent terms in a nonlinear manner, making them more suitable for bilinear Koopman models, which are more computationally expensive to train and deploy. Notably, the proposed scheme enables the formulation of linear models that achieve superior prediction performance compared to conventional bilinear models while remaining substantially more efficient. To realize this framework, we present two neural network architectures that construct Koopman embeddings from actuated or unactuated data, enabling flexible and efficient modeling across different tasks. Robustness is ensured through the integration of a linear Generalized Extended State Observer (GESO), which explicitly estimates disturbances and compensates for them in real time. The combined momentum-based Koopman and GESO framework is validated through comprehensive trajectory tracking simulations and experiments on robotic manipulators, demonstrating superior accuracy, robustness, and learning efficiency relative to state of the art alternatives.

A variety of recent methods guide large language model outputs via the inference-time addition of steering vectors to residual-stream or attention-head representations. In contrast, we propose to inject steering vectors directly into the query and value representation spaces within attention heads. We provide evidence that a greater portion of these spaces exhibit high linear discriminability of concepts --a key property motivating the use of steering vectors-- than attention head outputs. We analytically characterize the effect of our method, which we term DISentangled COmmunication (DISCO) Steering, on attention head outputs. Our analysis reveals that DISCO disentangles a strong but underutilized baseline, steering attention inputs, which implicitly modifies queries and values in a rigid manner. In contrast, DISCO's direct modulation of these components enables more granular control. We find that DISCO achieves superior performance over a number of steering vector baselines across multiple datasets on LLaMA 3.1 8B and Gemma 2 9B, with steering efficacy scoring up to 19.1% higher than the runner-up. Our results support the conclusion that the query and value spaces are powerful building blocks for steering vector methods.