The rise of the open-source RISC-V Instruction Set Architecture (ISA) has highlighted the critical need for a standardized framework for microarchitectural security evaluation, a gap that challenges the development of secure-by-design hardware. To address this, we present a novel virtual simulation platform, RAVEN, for evaluating the microarchitectural security of RISC-V processors. Built on the gem5 simulator, the proposed framework enables detailed modeling of cache hierarchies, integration of custom hardware performance counters, and fine-grained analysis of timing-based side-channel vulnerabilities. The platform is designed to support both attack emulation and automated detection, providing a flexible environment for architecture-level security evaluation. We validate the platform by implementing and analyzing five representative microarchitectural attacks: Spectre , Evict+Time , Evict+Spec+Time , Flush+Fault , and GhostRace . To detect these attacks, we integrate a machine learning-based detection pipeline that leverages hardware performance counter data to identify anomalous microarchitectural behavior. Our experimental results across multiple classifiers show high detection accuracy, demonstrating the platform’s effectiveness in distinguishing side-channel activity from benign execution. Overall, the extensible gem5-based platform provides a practical and scalable foundation for security research on the open RISC-V ISA, supporting both attack analysis and the development of secure-by-design hardware.
The dynamic data driven simulation paradigm continuously assimilates real-time data to improve system analysis and prediction. Operating dynamic data driven simulation in a distributed manner on edge platforms can significantly reduce latency and alleviate network bandwidth constraints. This article develops a general framework for distributed dynamic data driven simulation, specifically addressing highly nonlinear and non-Gaussian dynamics and hybrid states comprising continuous, discrete, and categorical variables. The framework operates on two interconnected levels: locally, a particle filter-based data assimilation methodology incorporates noisy measurements to estimate state distributions; globally, a decentralized, consensus-based fusion reconciles conflicting estimates across overlapping regions. To ensure communication efficiency, a hybrid belief compression and particle regeneration mechanism are designed to avoid the prohibitive exchange of raw particles. The framework is illustrated through a distributed microscopic traffic simulation case study featuring an accident-induced shockwave. The results show that the proposed framework can produce consistent hybrid state estimates while maintaining a low communication burden. In the reported traffic case, the sensitivity analysis further suggests robustness to microscopic sensor miss-counts, while also revealing a case-specific dependence on accurate spatial accumulation data.
This report presents the reproducibility evaluation for the paper “A Framework for Efficient Individual-Level Longitudinal Outcomes Estimation for Agent-Based Simulations.” The artifact was retrieved from a permanent Zenodo repository and evaluated on a Linux Ubuntu system using MATLAB R2025b. Our evaluation confirms that the proposed OAA and SPSA methods effectively reduce computational runtimes while maintaining the statistical integrity of the simulation results. The findings in terms of health outcomes and cost distributions are consistent with those reported by the authors. Consequently, the paper is awarded the Artifacts Available , Artifacts Evaluated—Reusable , and Results Validated - Results Reproduced badges.
Large language models (LLMs) have been shown to be capable of generating human-like agent behavior in diverse scenarios, making them useful building blocks for agent-based simulations. However, the substantial inference cost restricts the crowd sizes that can be tackled, and the opaque nature of LLM-based decision making raises reliability concerns. To address these issues, we propose the approach of Decision Function Distillation (DFD), which extracts strategies underlying the decision making of LLM agents in a rule-based and interpretable form. The final decision function is determined in an iterative process during which intermediate insights gathered from historical agent trajectories are refined and finally translated into commented code. In two variants of the approach, the intermediate insights are either generated directly as text based on few-shot examples or explicitly formalized into code snippets. Unlike black-box symbolic regression (SR), the gradual and transparent refinement process allows modelers to understand the strategies captured in the final commented decision function. We demonstrate DFD on agent-based crowd evacuation scenarios, showing that DFD outperforms both classical and a state-of-the-art LLM-based SR.
Using mixed methods approaches for problem-solving has a long history in Operations Research (OR) and Modelling and Simulation (M&S). Data Science (DS), with its strong alignment with the disciplines of Mathematics, Statistics and Computer Science, has experienced a surge in interest in recent decades and is now increasingly applied in business and management. Similarly, M&S, a sub-field of OR, has a long history of theoretical and practical work in the dynamic modelling of operational systems. Thus, hybrid models employing DS techniques such as supervised machine learning and reinforcement learning with M&S approaches like agent-based modelling and discrete-event simulation enable us to realise synergies associated with multiple methods; their combined application potentially goes beyond what could be possible by employing single techniques. Through a comprehensive survey of 117 researchers and practitioners, our work aims to identify the key challenges and opportunities in developing hybrid models. Our findings suggest that hybrid M&S-DS models can improve model accuracy, reduce computational costs, improve efficiency and potentially lead to improved decision-making. Our study advances the M&S knowledge base by broadening the methodological foundations that current and future researchers engage with, showcasing how combining traditional M&S approaches with emerging DS techniques deepens critical understanding of mixed methods approaches. By capturing both theoretical frameworks and practitioner insights, it supports the development of a more contemporary and comprehensive view of M&S practice.
Agent-based models (ABMs) are critical for projecting and responding to infectious disease outbreaks, yet the effectiveness of health interventions heavily relies on individual adherence, underscoring the importance of accurately representing human behavior. While some ABMs have incorporated behaviors, clear guidelines for leveraging the extensive literature on behavioral theories have been lacking. This study addresses this challenge by proposing a modeling framework that integrates classical behavioral theories and their psychological constructs, drawing upon well-established definitions and categorizations from the social sciences to guide operationalization. We present novel and existing modeling approaches to represent these constructs and their interactions with the environment and behavior. Furthermore, we discuss methods for determining an agent’s likelihood of adopting a behavior and how empirical data can be used for calibration and validation. To illustrate the application of the framework to disease-specific behaviors, we provide a modeling example. Ultimately, this proposed framework offers a roadmap for computational modelers to incorporate scientifically grounded behaviors in the design and evaluation of simulation models and interventions.
In this paper, we consider an agent-based modeling situation where individual-level stochastic outcomes for a subset of agents have to be estimated over time. This situation is commonly found in agent-based models of infectious disease transmission, where outcomes such as life years survived are accumulated over time as a stochastic process. For an ABM with a large agent cohort, this longitudinal outcome estimation process can incur substantial computational expense. We present two methods to alleviate this expense, both of which do not involve any curve-fitting; however, they involve one-time generation of repositories of outcome realizations stratified based on agent characteristics. The first method, the outcomes accumulation and allocation (OAA) approach, involves accumulation of outcomes between relevant changes in agent status as well as random sampling and allocation of outcomes from the repository at relevant points of change in agent status. The second method, the sample path-based sampling and allocation (SPSA) approach, eliminates the need for outcomes accumulation within the ABM execution by allocating outcomes based on the sample path experienced by an agent. Computational experiments with an ABM of the spread of hepatitis C virus in the Indian context indicate that runtimes under the OAA and SPSA approaches are approximately three-fifth and one-third of the benchmark accumulation-only outcomes estimation approach, while yielding comparable outcome distributions and their expectations. We also provide theoretical support for equivalence of outcome expectations between the OAA, SPSA and the benchmark accumulation-only approaches, and discuss the worst-case space complexity of repository generation for each method.
Simulating believable multi-agent behavior remains a significant challenge, particularly in endowing agents with intrinsic motivations and the capacity for complex, emergent social interactions. This article presents an integrated framework that addresses these challenges by synergizing generative agents, powered by Large Language Models (LLMs), with a rich semantic environment and sophisticated internal state models. The framework leverages scene semantics derived from Building Information Modeling (BIM) to facilitate meaningful agent-environment interactions, employs the OCEAN personality model to instill agent heterogeneity, and uses a dynamic emotion model to drive internal state evolution. A novel emotion-based neural A* algorithm is introduced to translate these internal states into plausible, non-optimal pathfinding decisions. Experimental results, including quantitative analysis and ablation studies, demonstrate the framework's capability to generate agents that exhibit (1) human-like path choices influenced by their dynamic emotional states, and (2) emergent social behaviors, such as forming unscripted relationships, driven by a persistent memory and reflection mechanism. This work contributes a significant step towards creating more autonomous and lifelike virtual agents for complex simulations. CCS Concepts: center dot Computing methodologies -> Modeling and simulation ; Multi-agent systems ; In
"Towards Standardizing Validation Practices in Agent-Based Modeling: A Hierarchical ABM Validation Framework" [1] proposes a novel validation framework for the family of simulation models known as agent-based models (ABMs), particularly popular in the social sciences. The framework is based on, first a clear definition of three crucial concepts in ABMs: calibration, verification, and validation, and then a review of 17 validation approaches. This replicated computations results report focuses on the prototypical tool implementation of such framework, Hierarchical ABM Validation (HAV). Considering the replicability material revised by the authors after an iteration, the software was straightforward to install and use, and the experimental results from the article could be reproduced in seconds on a standard laptop machine. The article receives the badges Artifacts Available, Artifacts Functional, Artifacts Evaluated-Reusable and Results Validated-Results Reproduced.
Agent-Based Modeling (ABM) has gained significant traction due to its ability to represent diverse agents' behaviors and interactions accurately. However, ensuring the reliability and widespread acceptance of ABM necessitates rigorous validation. Unfortunately, existing literature often lacks integration of representative validation methods into a cohesive framework, hindering standardized validation practices. This study aims at proposing a comprehensive and pragmatic validation framework. Initially, we clarify three fundamental concepts: calibration, verification, and validation. Subsequently, we review 17 distinct validation approaches and categorize them based on their data requirements and suitability for various simulation methodologies. Aligned with the ABM procedures, we introduce a Hierarchical ABM Validation (HAV) framework structured across three tiers: agent level, model level, and output level. Each tier recommends appropriate validation methods contingent upon data availability, enhancing the HAV's applicability across diverse modeling scenarios. Finally, we develop an accessible Python code package (hav) and provide two examples of validating a traffic model and a wealth model. These cases exemplify the HAV's implementation and underscore its efficacy in promoting standardized validation practices within ABM research.
Discrete-event (DE) systems are concurrent programs where components communicate via tagged events, where tags are drawn from a totally ordered set. Distributed DE (DDE) systems are DE systems where the components (reactors) communicate over networks. Most execution platforms require that for DDE systems with cycles, each cycle must contain at least one logical delay, where the tag of events is incremented. Some impose an even stronger constraint, that no component produces outputs with the same timestamp as a triggering input (the “lookahead” for the component must be greater than zero). Such restrictions, however, are not required by the elegant fixed-point semantics of DE. The only fundamental requirement is that the program be constructive, meaning it is free from causality cycles. In this article, we propose a way to coordinate the execution of DDE systems that can execute any constructive program, even one with zero-delay cycles (ZDC), facilitating the elegant programming of strongly consistent distributed real-time systems. The proposed coordination provides a formal model that exposes exactly the information that must be shared across networks for such execution to be possible. Our solution avoids speculative execution and rollback, making it suitable for situations that do not tolerate rollback, such as deployment (vs. simulation) of cyber-physical systems (CPS’s). We describe an extension to the coordination mechanisms in Lingua Franca, a recent DE-based coordination language, to support ZDC.
Stochastic simulation aims to compute output performance for complex models that lack analytical tractability. To ensure accurate prediction, the model needs to be calibrated and validated against real data. Conventional methods approach these tasks by assessing the model-data match via simple hypothesis tests or distance minimization in an ad hoc fashion, but they can encounter challenges arising from non-identifiability and high dimensionality. In this article, we investigate a framework to develop calibration schemes that satisfies rigorous frequentist statistical guarantees, via a basic notion that we call eligibility set designed to bypass non-identifiability via a set-based estimation. We investigate a feature extraction-then-aggregation approach to construct these sets that target at multivariate outputs. We demonstrate our methodology on several numerical examples, including an application to calibration of a limit order book market simulator (ABIDES).
Healthcare is a service-based system where patients and providers (clinicians and managers) are integral to the system's overall functioning. As a human-centric system at both the demand and supply sides, it is important to consider the behavioral aspects of the entities that make up the healthcare system and their effect on system performance. Modeling and Simulation (M&S) is widely used in healthcare to give stakeholders better insights into real-world problems; incorporating human behavior makes the models more realistic and leads to more informed decision-making. M&S for behavioral analysis (BA) seeks to understand and explore individuals' behavior and reactions to different interventions within healthcare systems. We reviewed 59 papers published between 1992 and 2023 to synthesize existing literature and ascertain its current development. Our findings show that in most studies, the impact of human behavior is either ignored, studied unsystematically, or treated as a second-order effect. The study contributes to the understanding of M&S methodologies for BA in healthcare.
We introduce a framework for constructing confidence intervals for the performance of a system as a function of a parameter, decision variable or system state, even when the system is not simulated at the particular parameter, decision variable or state. The proposed methods leverage observations from some other simulated model instances and known functional properties of the performance function being evaluated. The intervals, termed plausible intervals , deliver a desired coverage probability uniformly over all model instances as the minimum sample size at the simulated model instances increases, and they attain the strongest possible consistency from simulating a finite number of model instances. We illustrate the versatility and effectiveness of plausible intervals through two numerical experiments.
In delay-tolerant networks (DTNs) with uncertain contact plans, the communication episodes and their reliabilities are known a priori. To maximise the end-to-end delivery probability, a bounded network-wide number of message copies are allowed. The resulting multi-copy routing optimization problem is naturally modelled as a Markov decision process with distributed information. In this paper, we provide an in-depth comparison of three solution approaches: statistical model checking with scheduler sampling, the analytical RUCoP algorithm based on probabilistic model checking, and an implementation of concurrent Q-learning. We use an extensive benchmark set comprising random networks, scalable binomial topologies, and realistic ring-road low Earth orbit satellite networks. We evaluate the obtained message delivery probabilities as well as the computational effort. Our results show that all three approaches are suitable tools for obtaining reliable routes in DTN, and expose a tradeoff between scalability and solution quality.
Although high-fidelity network simulations have proven to be reliable and cost-effective tools to peer into architectural questions for high-performance computing (HPC) networks, they incur a high resource cost. The time spent in simulating a single millisecond of network traffic in the highest detail can take hours, even for static, well-behaved traffic patterns such as uniform random. Surrogate models offer a significant reduction in runtime, yet they cannot serve as complete replacements and should only be used when appropriate. Thus, there is a need for hybrid modeling, where high-fidelity simulation and surrogates run side-by-side. We present a surrogate model for HPC networks in which: packets bypass the network, while the network state is left untouched, i.e., suspended. To bypass the network, we use historical data to estimate the arrival time at which every packet should be scheduled at; to suspend the network, all in-flight packets are scheduled to arrive at their destinations, and are kept in the system to awaken as zombies when switching back to high-fidelity. Speedup for a hybrid model is relative to the proportion of surrogate to high-fidelity. This light-weight surrogate obtained up to 76x speedup. Keeping the zombies in the network showed an increase in the accuracy of the high-fidelity simulation on restart when compared to restarting the network from an empty state.
The artifact evaluated in this report is relevant to the article “Using (Not-so) Large Language Models to Generate Simulation Models in a Formal DSL: A Study on Reaction Networks”. The authors provided the code residing on a permanent repository, the instructions for building and executing the artifact are well-documented, as well as the dependencies needed and possible issues encountered during the reproducibility process. The process of running the experiments and generating data, plots, and table terminates correctly. The results could be reproduced. The article receives the badges Artifacts Available, Artifacts Evaluated—Reusable and Results Validated—Results Reproduced.
