检索结果-内蒙古大学图书馆

Conformal Tail Risk Control for Large Language Model Alignment

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Chen, Catherine Yu-Chi Shen, Jingyan Deng, Zhun Lei, Lihua Institute for Computational and Mathematical Engineering Stanford University United States Department of Industrial Engineering and Operations Research Columbia University United States Department of Computer Science University of North Carolina Chapel Hill United States Graduate School of Business Department of Statistics Stanford University United States

Recent developments in large language models (LLMs) have led to their widespread usage for various tasks. The prevalence of LLMs in society implores the assurance on the reliability of their performance. In particular, risk-sensitive applications demand meticulous attention to unexpectedly poor outcomes, i.e., tail events, for instance, toxic answers, humiliating language, and offensive outputs. Due to the costly nature of acquiring human annotations, general-purpose scoring models have been created to automate the process of quantifying these tail events. This phenomenon introduces potential human-machine misalignment between the respective scoring mechanisms. In this work, we present a lightweight calibration framework for blackbox models that ensures the alignment of humans and machines with provable guarantees. Our framework provides a rigorous approach to controlling any distortion risk measure that is characterized by a weighted average of quantiles of the loss incurred by the LLM with high confidence. The theoretical foundation of our method relies on the connection between conformal risk control and a traditional family of statistics, i.e., L-statistics. To demonstrate the utility of our framework, we conduct comprehensive experiments that address the issue of human-machine misalignment. © 2025, CC BY.

关键词： Risk assessment

Characterizations of Ni-loaded lignite char catalysts and their performance enhancements to catalytic steam gasification of coal

学校读者我要写书评

暂无评论

Journal of the Energy Institute 2022年 105卷 53-71页

作者： Tipo, Ronnachai Tippayawong, Nakorn Chaichana, Chatchawan Chimupala, Yothin Chaiklangmuang, Suparin Graduate Program in Industrial Chemistry Faculty of Science Chiang Mai University Chiang Mai50200 Thailand Department of Industrial Chemistry Faculty of Science Chiang Mai University Chiang Mai50200 Thailand Department of Mechanical Engineering Faculty of Engineering Chiang Mai University Chiang50200 Thailand Center of Excellence in Materials Science and Technology Faculty of Science Chiang Mai University Chiang Mai50200 Thailand Materials Science Research Center Faculty of Science Chiang Mai University Chiang Mai50200 Thailand

Catalyst characteristics such as specific surface area, porosity and active site play an essential role in catalyst design. Understanding of the characteristic functions in controlling catalyst activity is necessary for considering the catalyst application, especially the Ni-loaded lignite char catalyst, employed in catalytic steam gasification. In this work, the unmodified and five modified catalysts were used to study the effect of catalyst characteristics on their performance for catalytic steam gasification of coal in the two-stage fixed-bed reactor. The findings indicated that the fresh acid modified catalysts increased specific surface area and micropore formation, including the smaller Ni metal crystallite size, the decreasing of Ni metal nanoparticle size and the well-dispersed of Ni metal nanoparticles. Meanwhile, the fresh modified catalysts by the alkali and the combined alkali-acid treatments promoted mesopore formation. Enhancing catalytic steam gasification of coal, it was found that the mesoporosity of catalyst was the considerable initial characteristic for controlling the catalyst activity. The total gas yields of the mesoporous catalysts were in the range of 50.65–77.70 mmol/gVM,feed, while the total gas yields of the microporous catalysts were 43.82–55.34 mmol/gVM,feed. After catalyst testings, the specific surface areas and number of mesopores in the spent catalysts were increased by steam. Remarkably, these changing pore sizes affected the transformations of the isotherms (type I + II to type II) and the hysteresis loops (type H4 to type H3). Also, steam influenced the conversion of Ni metals to NiO, Ni(OH)2 and Ni2O3. For thermal property, all spent catalysts had high thermal stability. The modified catalysts could decrease carbon deposition more than the unmodified catalyst, which was an upside for catalyst reusability. © 2022 Energy Institute

关键词： Catalyst activity

Direct Teaching/Playback Method for the Contact Task of a Six-DOF Manipulator Using a Parallel-Wire-Type Teaching Device with a Remote Center Compliance Device 23

学校读者我要写书评

暂无评论

Direct Teaching/Playback Method for the Contact Task of a Si...

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Okochi, Tatsuya Sakai, Masao Morita, Yoshifumi Graduate School of Engineering Nagoya Institute of Technology Electrical and Mechanical Engineering Program Department of Engineering Gokiso-cho Aichi Nagoya466-8555 Japan Industrial Research Center Aichi Center for Industry and Science Technology Onda-cho 1-157-1 Aichi Kariya-shi448-0013 Japan Nagoya Institute of Technology Department of Electrical and Mechanical Engineering Gokiso-cho Showa-ku Aichi Nagoya466-8555 Japan

ISBN: (纸本)9788993215274

This study proposes a direct teaching method for the task of pressing an end effector attached to a robot manipulator against a workpiece with a constant force. When performing contact tasks, such as deburring or polishing using a manipulator, a passive compliance is required for the end effector to suppress excessive contact force and impact. Therefore, we developed an end effector that combines a 'parallel wire-type teaching device,' which enables direct teaching, and a compliance device. In the case of pressing a workpiece with a constant force, after teaching the robot's trajectory to trace the edge of the workpiece, the amount of pressing can be calculated from the stiffness of the end effector and the desired pressing force;the robot's trajectory can be edited to teach contact tasks. An experiment was conducted to press a straight trajectory with a constant force as the target value and evaluate the accuracy of force teaching. © 2023 ICROS.

关键词： End effectors

Solving the waste bin location problem with uncertain waste generation rate: a bi-objective robust optimization approach

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Rossit, Diego Bard, Jonathan CONICET 1253 Alem Avenue Bahía BlancaB8000CPB Argentina Graduate Program in Operations Research & Industrial Engineering University of Texas at Austin 204 E. Dean Keeton Street AustinTX78712 United States

An efficient Municipal solid waste (MSW) system is critical to modern cities in order to enhance sustainability and livability of urban life. With this aim, the planning phase of the MSW system should be carefully addressed by decision makers. However, planning success is dependent on many sources of uncertainty that can affect key parameters of the system, e.g., the waste generation rate in an urban area. With this in mind, this paper contributes with a robust optimization model to design the network of collection points (i.e., location and storage capacity), which are the first points of contact with the MSW system. A central feature of the model is a bi-objective function that aims at simultaneously minimizing the network costs of collection points and the required collection frequency to gather the accumulated waste (as a proxy of the collection cost). The value of the model is demonstrated by comparing its solutions with those obtained from its deterministic counterpart over a set of realistic instances considering different scenarios defined by different waste generation rates. The results show that the robust model finds competitive solutions in almost all cases investigated. An additional benefit of the model is that it allows the user to explore trade-offs between the two objectives. © 2024, CC BY-NC-ND.

关键词： Municipal solid waste

Decision-making methods for selecting the best strategy for Industry 4.0

学校读者我要写书评

暂无评论

International Journal of Manufacturing Technology and Management 2023年第5-6期37卷 538-562页

作者： Chonsawat, Nilubon Sopadang, Apichat Ouzrout, Yacine Graduate Program in Industrial Engineering Department of Industrial Engineering Chiang Mai University Chiang Mai50200 Thailand Department of Industrial Engineering Supply Chain Engineering Management Research Unit Chiang Mai University Chiang Mai50200 Thailand DISP Laboratory University Lumiere Lyon 2 160 Bd de l’Universite Bron69676 France

Emerging Industry 4.0 creates a critical challenge when an organisation decides to reform and adapt to developing technologies. The decision maker must be aware of the elements prior to implementing Industry 4.0 strategies and technologies. Following the research approach, this article presents the decision making framework for SMEs in Industry 4.0 implementation decisions. This research begins with Industry 4.0 aspects and readiness maturity levels that identify SME capability. This model is a hybrid multi-criteria method: the fuzzy DEMATEL was used to evaluate the direction of the aspect, and then the fuzzy best-worst method analysed the weighting of aspects. After that, the prioritisation of strategies was ranked. The result concluded the priority of SMEs to select the suitable strategy toward Industry 4.0. Moreover, human skills are the core organisational aspects that embrace adaptation. Finally, the research not only highlights the best strategies selection but also presents the human skills that support organisational implementation. Copyright © 2023 Inderscience Enterprises Ltd.

关键词： Decision making

Deep Learning-Based Residual Useful Lifetime Prediction for Assets with Uncertain Failure Modes

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Su, Yuqi Fang, Xiaolei Operations Research Graduate Program North Carolina State University 915 Partners Way RaleighNC27607 United States Edward P. Fitts Department of Industrial and Systems Engineering North Carolina State University 915 Partners Way RaleighNC27607 United States

industrial prognostics focuses on utilizing degradation signals to forecast and continually update the residual useful life of complex engineering systems. However, existing prognostic models for systems with multiple failure modes face several challenges in real-world applications, including overlapping degradation signals from multiple components, the presence of unlabeled historical data, and the similarity of signals across different failure modes. To tackle these issues, this research introduces two prognostic models that integrate the mixture (log)-location-scale distribution with deep learning. This integration facilitates the modeling of overlapping degradation signals, eliminates the need for explicit failure mode identification, and utilizes deep learning to capture complex nonlinear relationships between degradation signals and residual useful lifetimes. Numerical studies validate the superior performance of these proposed models compared to existing methods. © 2024, CC BY.

关键词： Failure modes

Blockchain: A Review from the Perspective of operations researchers

学校读者我要写书评

暂无评论

Blockchain: A Review from the Perspective of Operations Rese...

Simulation Winter Conference

作者： Hong Wan Kejun Li Yining Huang Department of Industrial and Systems Engineering North Carolina State University Raleigh NC USA Operations Research Graduate Program North Carolina State University Raleigh NC USA

ISBN: (纸本)9781665476621

Blockchain is a distributed append-only digital ledger. The technology has caught much attention since the emergence of cryptocurrency, and there is an increasing number of blockchain applications in various businesses. The concept, however, is still novel to many members of the simulation and operations research community. In this tutorial, we introduce the blockchain technology and review its frontier related research. There are exciting opportunities for researchers in simulation, system analysis, and data science.

关键词： Analytical models operations research Tutorials Data science Data models Blockchains Cryptocurrency

Atmospheric humidity regulates same-sex mating in Candida albicans through the trehalose and osmotic signaling pathways

学校读者我要写书评

暂无评论

Science China(Life Sciences) 2023年第8期66卷 1915-1929页

作者： Chao Li Li Tao Guobo Guan Zhangyue Guan Austin M.Perry Tianren Hu Jian Bing Ming Xu Clarissa J.Nobile Guanghua Huang Department of Infectious Diseases Huashan HospitalShanghai Institute of Infectious Disease and Biosecurity and State Key Laboratory of Genetic EngineeringSchool of Life SciencesFudan UniversityShanghai 200438China State Key Laboratory of Mycology Institute of MicrobiologyChinese Academy of SciencesBeijing 100101China Department of Molecular and Cell Biology University of CaliforniaMercedMerced CA 95343USA Quantitative and Systems Biology Graduate Program University of CaliforniaMercedMerced CA 95343USA Health Sciences Research Institute University of CaliforniaMercedMerced CA 95343USA Shanghai Engineering Research Center of Industrial Microorganisms Shanghai 200438China Shanghai Huashen Institute of Microbes and Infections Shanghai 200052China

Sexual reproduction is prevalent in eukaryotic organisms and plays a critical role in the evolution of new traits and in the generation of genetic *** factors often have a direct impact on the occurrence and frequency of sexual reproduction in *** regulatory effects of atmospheric relative humidity(RH)on sexual reproduction and pathogenesis in plant fungal pathogens and in soil fungi have been extensively ***,the knowledge of how RH regulates the lifecycles of human fungal pathogens is *** this study,we report that low atmospheric RH promotes the development of mating projections and same-sex(homothallic)mating in the human fungal pathogen Candida *** RH causes water loss in *** cells,which results in osmotic stress and the generation of intracellular reactive oxygen species(ROS)and *** water transporting aquaporin Aqy1,and the G-protein coupled receptor Gpr1 function as cell surface sensors of changes in atmospheric *** of the trehalose metabolic pathway by inactivating trehalose synthase or trehalase promotes same-sex mating in *** by increasing osmotic or ROS stresses,*** trehalose and ROS signal the Hog1-osmotic and Hsf1-Hsp90 signaling pathways to regulate the mating ***,therefore,propose that the cell surface sensors Aqy1 and Gpr1,intracellular trehalose and ROS,and the Hog1-osmotic and Hsf1-Hsp90 signaling pathways function coordinately to regulate sexual mating in response to low atmospheric RH conditions in ***.

关键词： atmospheric relative humidity sexual reproduction same-sex mating Candida albicans trehalose metabolism osmotic stress

A decision rule approach for two-stage data-driven distributionally robust optimization problems with random recourse

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Fan, Xiangyi Hanasusanto, Grani A. Graduate Program in Operations Research and Industrial Engineering The University of Texas Austin United States

We study two-stage stochastic optimization problems with random recourse, where the adaptive decisions are multiplied with the uncertain parameters in both the objective function and the constraints. To mitigate the computational intractability of infinite-dimensional optimization, we propose a scalable approximation scheme via piecewise linear and piecewise quadratic decision rules. We then develop a data-driven distributionally robust framework with two layers of robustness to address distributionally uncertainty. The emerging optimization problem can be reformulated as an exact copositive program, which admits tractable approximations in semidefinite programming. We design a decomposition algorithm where smaller-size semidefinite programs can be solved in parallel, which further reduces the runtime. Lastly, we establish the performance guarantees of the proposed scheme and demonstrate its effectiveness through numerical examples. Copyright © 2021, The Authors. All rights reserved.

关键词： Piecewise linear techniques