Autonomous driving evaluation requires simulation environments that closely replicate actual road conditions, including real-world sensory data and responsive feedback loops. However, many existing simulations need to predict waypoints along fixed routes on public datasets or synthetic photorealistic data, \ie, open-loop simulation usually lacks the ability to assess dynamic decision-making. While the recent efforts of closed-loop simulation offer feedback-driven environments, they cannot process visual sensor inputs or produce outputs that differ from real-world data. To address these challenges, we propose DrivingSphere, a realistic and closed-loop simulation framework. Its core idea is to build 4D world representation and generate real-life and controllable driving scenarios. In specific, our framework includes a Dynamic Environment Composition module that constructs a detailed 4D driving world with a format of occupancy equipping with static backgrounds and dynamic objects, and a Visual Scene Synthesis module that transforms this data into high-fidelity, multi-view video outputs, ensuring spatial and temporal consistency. By providing a dynamic and realistic simulation environment, DrivingSphere enables comprehensive testing and validation of autonomous driving algorithms, ultimately advancing the development of more reliable autonomous cars. The benchmark will be publicly released.

Reinforcement learning (RL) agents are powerful tools for managing power grids. They use large amounts of data to inform their actions and receive rewards or penalties as feedback to learn favorable responses for the system. Once trained, these agents can efficiently make decisions that would be too computationally complex for a human operator. This ability is especially valuable in decarbonizing power networks, where the demand for RL agents is increasing. These agents are well suited to control grid actions since the action space is constantly growing due to uncertainties in renewable generation, microgrid integration, and cybersecurity threats. To assess the efficacy of RL agents in response to an adverse grid event, we use the Grid2Op platform for agent training. We employ a proximal policy optimization (PPO) algorithm in conjunction with graph neural networks (GNNs). By simulating agents' responses to grid events, we assess their performance in avoiding grid failure for as long as possible. The performance of an agent is expressed concisely through its reward function, which helps the agent learn the most optimal ways to reconfigure a grid's topology amidst certain events. To model multi-actor scenarios that threaten modern power networks, particularly those resulting from cyberattacks, we integrate an opponent that acts iteratively against a given agent. This interplay between the RL agent and opponent is utilized in N-k contingency screening, providing a novel alternative to the traditional security assessment.

Donald Trump said on Saturday that Chris Wright, an oil and gas industry executive and a staunch defender of fossil fuel use, would be his pick to lead the US Department of Energy. Wright is the founder and CEO of Liberty Energy, an oilfield services firm based in Denver, Colorado. He is expected to support Trump's plan to maximize production of oil and gas and to seek ways to boost generation of electricity, demand for which is rising for the first time in decades. He is also likely to share Trump's opposition to global cooperation on fighting climate change. Wright has called climate change activists alarmist and has likened efforts by Democrats to combat global warming to Soviet-style communism.

This paper presents a framework for developing efficient and interpretable carbon-cycle emulators (CCEs) as part of climate emulators in Integrated Assessment Models, enabling economists to custom-build CCEs accurately calibrated to advanced climate science. We propose a generalized multi-reservoir linear box-model CCE that preserves key physical quantities and can be use-case tailored for specific use cases. Three CCEs are presented for illustration: the 3SR model (replicating DICE-2016), the 4PR model (including the land biosphere), and the 4PR-X model (accounting for dynamic land-use changes like deforestation that impact the reservoir's storage capacity). Evaluation of these models within the DICE framework shows that land-use changes in the 4PR-X model significantly impact atmospheric carbon and temperatures -- emphasizing the importance of using tailored climate emulators. By providing a transparent and flexible tool for policy analysis, our framework allows economists to assess the economic impacts of climate policies more accurately.

Modern industrial infrastructures rely heavily on Cyber-Physical Systems (CPS), but these are vulnerable to cyber-attacks with potentially catastrophic effects. To reduce these risks, anomaly detection methods based on physical invariants have been developed. However, these methods often require domain-specific expertise to manually define invariants, making them costly and difficult to scale. To address this limitation, we propose a novel approach to extract physical invariants from CPS testbeds for anomaly detection. Our insight is that CPS design documentation often contains semantically rich descriptions of physical procedures, which can profile inter-correlated dynamics among system components. Leveraging the built-in physics and engineering knowledge of recent generative AI models, we aim to automate this traditionally manual process, improving scalability and reducing costs. This work focuses on designing and optimizing a Retrieval-Augmented-Generation (RAG) workflow with a customized prompting system tailored for CPS documentation, enabling accurate extraction of semantic information and inference of physical invariants from complex, multimodal content. Then, rather than directly applying the inferred invariants for anomaly detection, we introduce an innovative statistics-based learning approach that integrates these invariants into the training dataset. This method addresses limitations such as hallucination and concept drift, enhancing the reliability of the model. We evaluate our approach on real-world public CPS security dataset which contains 86 data points and 58 attacking cases. The results show that our approach achieves a high precision of 0.923, accurately detecting anomalies while minimizing false alarms.

Remote Sensing Visual Question Answering (RSVQA) is a challenging task that involves interpreting complex satellite imagery to answer natural language questions. Traditional approaches often rely on separate visual feature extractors and language processing models, which can be computationally intensive and limited in their ability to handle open-ended questions. In this paper, we propose a novel method that leverages a generative Large Vision-Language Model (LVLM) to streamline the RSVQA process. Our approach consists of a two-step training strategy: domain-adaptive pretraining and prompt-based finetuning. This method enables the LVLM to generate natural language answers by conditioning on both visual and textual inputs, without the need for predefined answer categories. We evaluate our model on the RSVQAxBEN dataset, demonstrating superior performance compared to state-of-the-art baselines. Additionally, a human evaluation study shows that our method produces answers that are more accurate, relevant, and fluent. The results highlight the potential of generative LVLMs in advancing the field of remote sensing analysis.

As core thermal power generation equipment, steam turbines incur significant expenses and adverse effects on operation when facing interruptions like downtime, maintenance, and damage. Accurate anomaly detection is the prerequisite for ensuring the safe and stable operation of steam turbines. However, challenges in steam turbine anomaly detection, including inherent anomalies, lack of temporal information analysis, and high-dimensional data complexity, limit the effectiveness of existing methods. To address these challenges, we proposed an Enhanced Long Short-Term Memory Variational Autoencoder using Deep Advanced Features and Gaussian Mixture Model (ELSTMVAE-DAF-GMM) for precise unsupervised anomaly detection in unlabeled datasets. Specifically, LSTMVAE, integrating LSTM with VAE, was used to project high-dimensional time-series data to a low-dimensional phase space. The Deep Autoencoder-Local Outlier Factor (DAE-LOF) sample selection mechanism was used to eliminate inherent anomalies during training, further improving the model's precision and reliability. The novel deep advanced features (DAF) hybridize latent embeddings and reconstruction discrepancies from the LSTMVAE model and provide a more comprehensive data representation within a continuous and structured phase space, significantly enhancing anomaly detection by synergizing temporal dynamics with data pattern variations. These DAF were incorporated into GMM to ensure robust and effective unsupervised anomaly detection. We utilized real operating data from industry steam turbines and conducted both comparison and ablation experiments, demonstrating superior anomaly detection outcomes characterized by high accuracy and minimal false alarm rates compared with existing methods.

We discovered the underlying physics in Next-token Prediction (NTP). We identified the law of information conservation within NTP and proposed the First Law of Information Capacity (IC-1), demonstrating that the essence of intelligence emergence in auto-regressive models is fundamentally a process of information transfer. We also introduced Landauer's Principle into NTP, formulating the Second Law of Information Capacity (IC-2), which establishes the relationship between auto-regressive model training and energy consumption. Additionally, we presented several corollaries, which hold practical significance for production practices. Finally, we demonstrate the consistency between the Law of Information Capacity and the Scaling Law for Neural Language Models, the Knowledge Capacity Scaling Laws, and the Scaling Laws for Precision.

In the endeavor to make autonomous robots take actions, task planning is a major challenge that requires translating high-level task descriptions into long-horizon action sequences. Despite recent advances in language model agents, they remain prone to planning errors and limited in their ability to plan ahead. To address these limitations in robotic planning, we advocate a self-refining scheme that iteratively refines a draft plan until an equilibrium is reached. Remarkably, this process can be optimized end-to-end from an analytical perspective without the need to curate additional verifiers or reward models, allowing us to train self-refining planners in a simple supervised learning fashion. Meanwhile, a nested equilibrium sequence modeling procedure is devised for efficient closed-loop planning that incorporates useful feedback from the environment (or an internal world model). Our method is evaluated on the VirtualHome-Env benchmark, showing advanced performance with better scaling for inference computation. Based on their extensive world knowledge, LLM agents seem close to autonomously performing robotic tasks, such as in household scenarios. However, growing evidence shows that existing LLM agents struggle with task planning (Kaelbling & Lozano-Pérez, 2011) that decomposes a high-level task into mid-level actions. While this problem requires long-horizon planning as well as consideration of environmental feedback, LLMs are often limited by: (1) unidirectional dependency: due to autoregressive generation, previous tokens cannot attend to future tokens, resulting in limited ability to plan ahead (Wu et al., 2024a); (2) lack of error correction for existing outputs, unless with a heavy system 2; (3) fixed forward process hindering the allocation of more inference computation to further improve planning performance. These inherent limitations of LLMs lead to inefficiency in the closed-loop long-horizon robotic planning. To address above challenges of LLM planners in closed-loop long-horizon planning, we advocate the approach of self-refinement (Welleck et al., 2023; Shinn et al., 2023; Kim et al., 2023; Madaan et al., 2023) that iteratively improves a previously generated plan. The reasons behind are threefold: (1) bidirectional dependency: since the output is conditioned on a previous draft plan, it can attend to all tokens in the plan (from an old version), thus improving its ability to plan ahead; (2) internal error correction which allows implicit self-correction in a forward pass without an explicit, heavy system 2; (3) dynamic computation allocation by iterating through a self-refinement process until convergence.

Accurate forecasts of distributed solar generation are necessary to reduce negative impacts resulting from the increased uptake of distributed solar photovoltaic (PV) systems. However, the high variability of solar generation over short time intervals (seconds to minutes) caused by cloud movement makes this forecasting task difficult. To address this, using cloud images, which capture the second-to-second changes in cloud cover affecting solar generation, has shown promise. Recently, deep neural networks with "attention" that focus on important regions of an image have been applied with success in many computer vision applications. However, their use for forecasting cloud movement has not yet been extensively explored. In this work, we propose an attention-based convolutional long short-term memory network to forecast cloud movement and apply an existing self-attention-based method previously proposed for video prediction to forecast cloud movement. We investigate and discuss the impact of cloud forecasts from attention-based methods towards forecasting distributed solar generation, compared to cloud forecasts from non-attention-based methods. We further provide insights into the different solar forecast performances that can be achieved for high and low altitude clouds. We find that for clouds at high altitudes, the cloud predictions obtained using attention-based methods result in solar forecast skill score improvements of 5.86% or more compared to non-attention-based methods.