With the rapid development of artificial intelligence, data-driven methods effectively overcome limitations in traditional portfolio optimization. Conventional models primarily employ long-only mechanisms, excluding highly correlated assets to diversify risk. However, incorporating short-selling enables low-risk arbitrage through hedging correlated assets. This paper constructs a Deep Reinforcement Learning (DRL) portfolio management framework with short-selling mechanisms conforming to actual trading rules, exploring strategies for excess returns in China's A-share market. Key innovations include: (1) Development of a comprehensive short-selling mechanism in continuous trading that accounts for dynamic evolution of transactions across time periods; (2) Design of a long-short optimization framework integrating deep neural networks for processing multi-dimensional financial time series with mean Sharpe ratio reward functions. Empirical results show the DRL model with short-selling demonstrates significant optimization capabilities, achieving consistent positive returns during backtesting periods. Compared to traditional approaches, this model delivers superior risk-adjusted returns while reducing maximum drawdown. From an allocation perspective, the DRL model establishes a robust investment style, enhancing defensive capabilities through strategic avoidance of underperforming assets and balanced capital allocation. This research contributes to portfolio theory while providing novel methodologies for quantitative investment practice.

Our approach integrates state estimation and dynamics modeling under a consistent architecture and training paradigm. Our diffusion-based perception model generates cloth states from partial observations, and the diffusion-based dynamics model generates physically plausible future states conditioned on action sequences, enabling robust model-based control. Our work demonstrates the potential of diffusion models in state estimation and dynamics modeling for manipulation tasks involving partial observability and complex dynamics. Abstract--Manipulating deformable objects like cloth is challenging states given the current state and robot actions. Leveraging a due to their complex dynamics, near-infinite degrees of transformer-based diffusion model, our method achieves highfidelity freedom, and frequent self-occlusions, which complicate state state reconstruction while reducing long-horizon dynamics estimation and dynamics modeling. Prior work has struggled with prediction errors by an order of magnitude compared to robust cloth state estimation, while dynamics models, primarily GNN-based approaches. Integrated with model-predictive control based on Graph Neural Networks (GNNs), are limited by their (MPC), our framework successfully executes cloth folding on a locality. Inspired by recent advances in generative models, we real robotic system, demonstrating the potential of generative hypothesize that these expressive models can effectively capture models for manipulation tasks with partial observability and intricate cloth configurations and deformation patterns from complex dynamics.

This paper introduces an innovative state estimator, MUSE (MUlti-sensor State Estimator), designed to enhance state estimation's accuracy and real-time performance in quadruped robot navigation. The proposed state estimator builds upon our previous work presented in [1]. It integrates data from a range of onboard sensors, including IMUs, encoders, cameras, and LiDARs, to deliver a comprehensive and reliable estimation of the robot's pose and motion, even in slippery scenarios. We tested MUSE on a Unitree Aliengo robot, successfully closing the locomotion control loop in difficult scenarios, including slippery and uneven terrain. Benchmarking against Pronto [2] and VILENS [3] showed 67.6% and 26.7% reductions in translational errors, respectively. Additionally, MUSE outperformed DLIO [4], a LiDAR-inertial odometry system in rotational errors and frequency, while the proprioceptive version of MUSE (P-MUSE) outperformed TSIF [5], with a 45.9% reduction in absolute trajectory error (ATE).

A nonparametric adaptive crane control system is proposed where the crane payload tracks a desired trajectory with feedback from the payload position. The payload motion is controlled with the position of the crane tip using partial feedback linearization. This is made possible by introducing a novel model structure given in Cartesian coordinates. This Cartesian model structure makes it possible to implement a nonparametric adaptive controller which cancels disturbances by approximating the effects of unknown disturbance forces and structurally unknown dynamics in a reproducing kernel Hilbert space (RKHS). It is shown that the nonparametric adaptive controller leads to uniformly ultimately bounded errors in the presence of unknown forces and unmodeled dynamics. Moreover, it is shown that the Cartesian formulation has certain advantages in payload tracking control also in the non-adaptive case. The performance of the nonparametric adaptive controller is validated in simulation and experiments with good results.

In this letter, we demonstrate that previously proposed improved state parameterizations for soft and continuum robots are specific cases of Clarke coordinates. By explicitly deriving these improved parameterizations from a generalized Clarke transformation matrix, we unify various approaches into one comprehensive mathematical framework. This unified representation provides clarity regarding their relationships and generalizes them beyond existing constraints, including arbitrary joint numbers, joint distributions, and underlying modeling assumptions. This unification consolidates prior insights and establishes Clarke coordinates as a foundational tool, enabling systematic knowledge transfer across different subfields within soft and continuum robotics.

With the rapid evolution of Large Language Models (LLMs) and their large-scale experimentation in cloud-computing spaces, the challenge of guaranteeing their security and efficiency in a failure scenario has become a main issue. To ensure the reliability and availability of large-scale language models in cloud computing scenarios, such as frequent resource failures, network problems, and computational overheads, this study proposes a novel adaptive fault tolerance mechanism. It builds upon known fault-tolerant mechanisms, such as checkpointing, redundancy, and state transposition, introducing dynamic resource allocation and prediction of failure based on real-time performance metrics. The hybrid model integrates data driven deep learning-based anomaly detection technique underlining the contribution of cloud orchestration middleware for predictive prevention of system failures. Additionally, the model integrates adaptive checkpointing and recovery strategies that dynamically adapt according to load and system state to minimize the influence on the performance of the model and minimize downtime. The experimental results demonstrate that the designed model considerably enhances the fault tolerance in large-scale cloud surroundings, and decreases the system downtime by $\mathbf{30\%}$, and has a better modeling availability than the classical fault tolerance mechanism.

End-to-end autonomous driving research currently faces a critical challenge in bridging the gap between open-loop training and closed-loop deployment. Current approaches are trained to predict trajectories in an open-loop environment, which struggle with quick reactions to other agents in closed-loop environments and risk generating kinematically infeasible plans due to the gap between open-loop training and closed-loop driving. In this paper, we introduce Hydra-NeXt, a novel multi-branch planning framework that unifies trajectory prediction, control prediction, and a trajectory refinement network in one model. Unlike current open-loop trajectory prediction models that only handle general-case planning, Hydra-NeXt further utilizes a control decoder to focus on short-term actions, which enables faster responses to dynamic situations and reactive agents. Moreover, we propose the Trajectory Refinement module to augment and refine the planning decisions by effectively adhering to kinematic constraints in closed-loop environments. This unified approach bridges the gap between open-loop training and closed-loop driving, demonstrating superior performance of 65.89 Driving Score (DS) and 48.20% Success Rate (SR) on the Bench2Drive dataset without relying on external experts for data collection. Hydra-NeXt surpasses the previous state-of-the-art by 22.98 DS and 17.49 SR, marking a significant advancement in autonomous driving. Code will be available at https://github.com/woxihuanjiangguo/Hydra-NeXt.

Ensuring safe and effective use of AI requires understanding and anticipating its performance on novel tasks, from advanced scientific challenges to transformed workplace activities. So far, benchmarking has guided progress in AI, but it has offered limited explanatory and predictive power for general-purpose AI systems, given the low transferability across diverse tasks. In this paper, we introduce general scales for AI evaluation that can explain what common AI benchmarks really measure, extract ability profiles of AI systems, and predict their performance for new task instances, in- and out-of-distribution. Our fully-automated methodology builds on 18 newly-crafted rubrics that place instance demands on general scales that do not saturate. Illustrated for 15 large language models and 63 tasks, high explanatory power is unleashed from inspecting the demand and ability profiles, bringing insights on the sensitivity and specificity exhibited by different benchmarks, and how knowledge, metacognition and reasoning are affected by model size, chain-of-thought and distillation. Surprisingly, high predictive power at the instance level becomes possible using these demand levels, providing superior estimates over black-box baseline predictors based on embeddings or finetuning, especially in out-of-distribution settings (new tasks and new benchmarks). The scales, rubrics, battery, techniques and results presented here represent a major step for AI evaluation, underpinning the reliable deployment of AI in the years ahead. (Collaborative platform: https://kinds-of-intelligence-cfi.github.io/ADELE.)

Despite their effectiveness, models in time series classification necessitate transformers are computationally expensive, effective optimization strategies for energyefficient making them less viable for real-time and edge-based deployment. Our study presents a systematic applications due to their high energy consumption investigation of optimization techniques, focusing and memory footprint. Furthermore, the growing on structured pruning and quantization methods carbon footprint associated with transformer training for transformer architectures. Through extensive and inference has raised significant concerns regarding experimentation on three distinct datasets (RefrigerationDevices, sustainability and deployment feasibility in ElectricDevices, and PLAID), we resource-constrained environments [22]. With the exponentially quantitatively evaluate model performance and energy increasing usage of AI, the carbon footprint efficiency across different transformer configurations. of massive models has been a topic of increasing Our experimental results demonstrate that worry. At the same time, deep learning model's runaway static quantization reduces energy consumption by scaling, such as huge transformers and diffusion 29.14% while maintaining classification performance, models, has induced unmatched computation costs, and L1 pruning achieves a 63% improvement in inference which are enormous and require a lot of power to operate speed with minimal accuracy degradation. Our and train, directly contributing to the increase findings provide valuable insights into the effectiveness in global carbon emissions.

High energy consumption remains a key challenge in high-performance computing (HPC) systems, which often feature hundreds or thousands of nodes drawing substantial power even in idle or standby modes. Although powering down unused nodes can improve energy efficiency, choosing the wrong time to do so can degrade quality of service by delaying job execution. Machine learning, in particular reinforcement learning (RL), has shown promise in determining optimal times to switch nodes on or off. In this study, we enhance the performance of a deep reinforcement learning (DRL) agent for HPC power management by integrating curriculum learning (CL), a training approach that introduces tasks with gradually increasing difficulty. Using the Batsim-py simulation framework, we compare the proposed CL-based agent to both a baseline DRL method (without CL) and the conventional fixed-time timeout strategy. Experimental results confirm that an easy-to-hard curriculum outperforms other training orders in terms of reducing wasted energy usage. The best agent achieves a 3.73% energy reduction over the baseline DRL method and a 4.66% improvement compared to the best timeout configuration (shutdown every 15 minutes of idle time). In addition, it reduces average job waiting time by 9.24% and maintains a higher job-filling rate, indicating more effective resource utilization. Sensitivity tests across various switch-on durations, power levels, and cluster sizes further reveal the agent's adaptability to changing system parameters without retraining. These findings demonstrate that curriculum learning can significantly improve DRL-based power management in HPC, balancing energy savings, quality of service, and robustness to diverse configurations.