The serial model of Epstein et al. （1981） is demonstrated to predict successfully the overall voidage of the mixed layer of incompletely segregated binary and ternary mixtures of liquid-fluidized solids that differ ...The serial model of Epstein et al. （1981） is demonstrated to predict successfully the overall voidage of the mixed layer of incompletely segregated binary and ternary mixtures of liquid-fluidized solids that differ only in shape. Some speculations are then offered on the conditions for the success and failure of this model. Finally, methods proposed in the literature for estimating the deviations from the serial model for a binary-solids bed, which manifest themselves as a bed contraction, are discussed, with emphasis on packing models.展开更多
The ordered event model is improved to describe serial production lines with limited buffer sizes. The improved model hasthe same computational burden as the original one and can be constrUcted directly according to g...The ordered event model is improved to describe serial production lines with limited buffer sizes. The improved model hasthe same computational burden as the original one and can be constrUcted directly according to given principles. Several simulationexamples are cited to verify this improved model. Extensions and open problems are also indicated. By means of this new model, serialProduotion lines with limitations of resources can be stUdied analytically.展开更多
According to the deficiency of the present model of pneumatic artificialmuscles (PAM), a serial model is built up based on the PAM's essential working principle with theelastic theory, it is validated by the quasi...According to the deficiency of the present model of pneumatic artificialmuscles (PAM), a serial model is built up based on the PAM's essential working principle with theelastic theory, it is validated by the quasi-static and dynamic experiment results, which are gainedfrom two experiment systems. The experiment results and the simulation results illustrate that theserial model has made a great success compared with Chou's model, which can describe the forcecharacteristics of PAM more precisely. A compensation item considering the braid's elasticity andthe coulomb damp is attached to the serial model based on the analysis of the experiment results.The dynamic experiment proves that the viscous damp of the PAM could be ignored in order to simplifythe model of PAM. Finally, an improved serial model of PAM is obtained.展开更多
大语言模型(Large Language Models,LLMs)凭借其庞大的参数规模和强大的语义表达能力,在自然语言处理、计算机视觉等领域取得突破性进展,并逐渐成为智能系统的关键基础.然而,随着模型轻量化、本地化定制及场景专用化需求持续增强,面向...大语言模型(Large Language Models,LLMs)凭借其庞大的参数规模和强大的语义表达能力,在自然语言处理、计算机视觉等领域取得突破性进展,并逐渐成为智能系统的关键基础.然而,随着模型轻量化、本地化定制及场景专用化需求持续增强,面向特定任务开发的专有化模型快速涌现.这类模型通常在局部领域具备能力优势,但难以独立覆盖多任务、多领域的复杂推理需求,从而推动了多模型协作推理的研究.现有研究多侧重于模型融合或单一协作范式,难以充分挖掘各模型间的优势互补潜力,且在协作结构和路径机制方面缺乏系统性的探索.为此,本文提出一种面向模型互联场景的多模型协作结构推理方法,构建了由线性链式结构向多路径组合结构演进的协作推理体系.在基础协作层面,设计了串联推理(Serial Inference,SI)与并联推理(Parallel Inference,PI)两种核心范式,分别通过阶段式信息传递与多模型并行处理增强推理过程中的语义收敛性与信息覆盖度.在此基础上,进一步从协作范式层面提出了“先串后并”(Serial-to-Parallel,S2P)与“先并后串”(Parallel-to-Serial,P2S)两种组合策略,实现协作路径在深度与广度之间的动态调度,拓展了多模型协作的结构表达能力与推理能力边界.本文在数学推理、知识理解和符号推理三类典型任务上搭建了系统实验框架,对四类协作策略进行了全面评估.实验结果表明,四类协作策略相较于单模型推理在平均准确率上分别提升了24.33、16.66、26.66和25.33个百分点.进一步分析发现,组合协作策略在融合串联与并联结构优势的同时,能够有效压缩整体推理时延,并在相较于最优单模型可接受的时延增量条件下,实现了更高的推理准确率,展现出在多任务场景下更优的性能-效率的权衡.此外,本文还系统分析了不同模型路径配置在协作过程中的表现差异,为多模型组网结构设计、协作机制优化及大规模模型互联体系的构建提供了理论依据与实证支撑.展开更多
文摘The serial model of Epstein et al. （1981） is demonstrated to predict successfully the overall voidage of the mixed layer of incompletely segregated binary and ternary mixtures of liquid-fluidized solids that differ only in shape. Some speculations are then offered on the conditions for the success and failure of this model. Finally, methods proposed in the literature for estimating the deviations from the serial model for a binary-solids bed, which manifest themselves as a bed contraction, are discussed, with emphasis on packing models.
文摘The ordered event model is improved to describe serial production lines with limited buffer sizes. The improved model hasthe same computational burden as the original one and can be constrUcted directly according to given principles. Several simulationexamples are cited to verify this improved model. Extensions and open problems are also indicated. By means of this new model, serialProduotion lines with limitations of resources can be stUdied analytically.
文摘According to the deficiency of the present model of pneumatic artificialmuscles (PAM), a serial model is built up based on the PAM's essential working principle with theelastic theory, it is validated by the quasi-static and dynamic experiment results, which are gainedfrom two experiment systems. The experiment results and the simulation results illustrate that theserial model has made a great success compared with Chou's model, which can describe the forcecharacteristics of PAM more precisely. A compensation item considering the braid's elasticity andthe coulomb damp is attached to the serial model based on the analysis of the experiment results.The dynamic experiment proves that the viscous damp of the PAM could be ignored in order to simplifythe model of PAM. Finally, an improved serial model of PAM is obtained.
文摘大语言模型(Large Language Models,LLMs)凭借其庞大的参数规模和强大的语义表达能力,在自然语言处理、计算机视觉等领域取得突破性进展,并逐渐成为智能系统的关键基础.然而,随着模型轻量化、本地化定制及场景专用化需求持续增强,面向特定任务开发的专有化模型快速涌现.这类模型通常在局部领域具备能力优势,但难以独立覆盖多任务、多领域的复杂推理需求,从而推动了多模型协作推理的研究.现有研究多侧重于模型融合或单一协作范式,难以充分挖掘各模型间的优势互补潜力,且在协作结构和路径机制方面缺乏系统性的探索.为此,本文提出一种面向模型互联场景的多模型协作结构推理方法,构建了由线性链式结构向多路径组合结构演进的协作推理体系.在基础协作层面,设计了串联推理(Serial Inference,SI)与并联推理(Parallel Inference,PI)两种核心范式,分别通过阶段式信息传递与多模型并行处理增强推理过程中的语义收敛性与信息覆盖度.在此基础上,进一步从协作范式层面提出了“先串后并”(Serial-to-Parallel,S2P)与“先并后串”(Parallel-to-Serial,P2S)两种组合策略,实现协作路径在深度与广度之间的动态调度,拓展了多模型协作的结构表达能力与推理能力边界.本文在数学推理、知识理解和符号推理三类典型任务上搭建了系统实验框架,对四类协作策略进行了全面评估.实验结果表明,四类协作策略相较于单模型推理在平均准确率上分别提升了24.33、16.66、26.66和25.33个百分点.进一步分析发现,组合协作策略在融合串联与并联结构优势的同时,能够有效压缩整体推理时延,并在相较于最优单模型可接受的时延增量条件下,实现了更高的推理准确率,展现出在多任务场景下更优的性能-效率的权衡.此外,本文还系统分析了不同模型路径配置在协作过程中的表现差异,为多模型组网结构设计、协作机制优化及大规模模型互联体系的构建提供了理论依据与实证支撑.
