cost management

研究通过FWBench工具评估了语言模型在成本限制下选择和使用时间序列预测进行决策的能力，测试了包括小型语言模型在内的十种配置。

This study addresses the finite-horizon budget allocation problem under non-stationary changes in return efficiency by formulating it as a closed-loop economic control problem. The authors employ a receding-horizon model predictive control (MPC) approach to dynamically optimize budget allocation, accounting for execution noise and operational constraints. Through comparison with reactive strategies, the research demonstrates that non-stationarity alone is insufficient for MPC to outperform reactive methods; MPC achieves significant and sustained superiority only when the return efficiency exhibits predictable structures that the model can effectively capture, thereby enabling advantageous intertemporal trade-offs. In contrast, under scenarios of random drift or stationarity, MPC offers no notable performance advantage over reactive approaches.

This study addresses how to dynamically allocate discretionary authority to street-level bureaucrats under resource and operational constraints, balancing policy consistency with case-specific optimization. Discretion is modeled as a dynamic budget allocation problem, and the authors propose a threshold-based rule that depends on time and remaining budget. Theoretical analysis reveals that, within location-scale distribution families, the optimal rate of discretion use depends solely on the shape of the distribution—such as tail thickness—and is invariant to the scale of benefits, reflecting a form of behavioral invariance termed “policy personality.” Integrating stochastic dynamic programming, threshold policy analysis, and empirical data from a homeless services system, the study demonstrates that discretionary behavior is significantly shaped by weekly work rhythms, weekend service interruptions, and short-term housing capacity, confirming that treating discretion as a budgetable resource can simultaneously enhance procedural fairness and social welfare.

Existing economic policy modeling suffers from fragmented instrument inventories, inconsistent classification criteria, and weak interpretability of policy combinations. To address these issues, this paper introduces the first taxonomy-based, hierarchical tree-structured framework for classifying economic policies. Grounded in economic theory, the framework systematically synthesizes and abstracts policy instruments, achieving exhaustive coverage and structured representation of all government-implementable macroeconomic levers. Compared with prior approaches, it enables multi-dimensional policy mapping, enhances transparency of model assumptions, and provides standardized inputs for macroeconomic simulation. Its core contributions are threefold: (1) a comprehensive, hierarchically organized policy classification system; (2) improved logical consistency and interpretability in policy package design; and (3) support for targeted, indicator-specific modeling—e.g., inflation, employment, or growth—and rigorous counterfactual analysis.

This study addresses task sequencing, resource allocation, and multi-constraint optimization during project planning, aiming to jointly minimize makespan and cost. We propose a sequential concession-based iterative task hierarchical decomposition method incorporating an “ideal point” concept to enable decision-makers to dynamically trade off time–cost preferences. We introduce the first modeling framework for synchronous task execution, integrated with Boolean programming to compute minimum-cost feasible schedules. Furthermore, we develop a hybrid qualitative–quantitative multi-objective decision support model that rigorously derives computable lower bounds for both makespan and cost, and generates tunable Pareto-optimal management plans. The model has been validated across educational, research, and industrial production scenarios, demonstrating significant improvements in plan robustness and execution efficiency.

This study addresses how agents can efficiently allocate external state observation resources under a shared budget. We propose BudgetPM, the first framework to formally define resource allocation in this setting. By leveraging a logistic scorer and full-episode hindsight distillation to train a lightweight policy network, combined with a hard-budget executor, our approach dynamically optimizes the timing of external inspections intended for memory storage. Furthermore, we devise differentiated strategies distinguishing between capacity-sufficient and capacity-scarce scenarios. Experimental results demonstrate that our method reduces observation overhead by 42–54% while maintaining near-perfect task performance. Under severely constrained budgets, it achieves substantial improvements in F1 score alongside a 16–33% reduction in observation volume.

本文提出内部外部性概念，解释了组织内部机制如何导致努力相互抵消，产生结构性低效，并建议先解决结构问题再优化资源配置。

Scheduling activities in business processes can improve efficiency (e.g., reduce makespan), but is challenging because the exact sequence of activities required to complete a case is often uncertain due to decisions based on data that emerges during execution. Nevertheless, probabilistic information regarding such decisions can often be estimated or derived from historical execution logs, and can help anticipate which execution paths are likely to lead to successful completion. Planning with particular execution paths affects feasibility, i.e., the probability of successful completion, and the expected number of superfluous activities that are planned but never executed. We frame the problem as a chance-constrained optimization problem and present two formulations: A decomposed approach with two stages, a planning stage that minimizes the expected number of superfluous activities subject to a feasibility constraint, and a scheduling stage that minimizes the makespan over the planned activities; and an integrated approach that combines planning and scheduling into a single formulation. Evaluation on two real-world and one synthetic dataset shows that the integrated approach yields superior makespans but is intractable at scale, while the decomposed approach scales to large settings.

This study addresses the challenge of quantifying the complexity and cost induced by external requirement changes when detailed knowledge of a system’s internal logic is unavailable. To this end, the authors propose a black-box assessment method based on a directed graph of component coupling. By analyzing component interfaces and integrating multi-view modeling—graphical, algebraic, and tabular—the approach uniquely links interface characteristics to cost factors, enabling computable bounded estimates of change-induced complexity and associated costs. The method was validated through a large-scale integration case in a retail banking platform, demonstrating its effectiveness and providing architects and operations teams with actionable, quantitative insights for system design and maintenance.

This work addresses the frequent conflation of technical debt (a stock liability) and stochastic tax (a flow burden) in AI agent systems, despite their fundamentally distinct natures and substantial implications for governance and operational cost assessment. The study proposes a structured framework that enables the independent quantification and joint analysis of these two constructs through modeling, operational data analysis, and simulation. To facilitate practical adoption, the framework is accompanied by an interactive dashboard and spreadsheet-based tools. Empirical validation in real-world business contexts—such as accounts payable—demonstrates that the approach effectively supports cost control and governance decisions in AI systems, offering practitioners actionable, quantifiable, and visualizable methods for managing systemic liabilities.