arXiv subjects

Holger Giese

Publications and source records attributed to Holger Giese.

At least 19 recordsLinked to original sources

arXiv2026

Interference-Aware Cross-Application Placement: A Multi-Objective Optimization Approach for Microservice Clusters

In modern cloud architectures, multiple applications often run within the same clustered environment, sharing underlying resources. This resource sharing can cause interference among applications, leading to degraded latency and reduced system stability. As containerized microservices become increasingly central to cloud-native applications, their performance can suffer from complex interference scenarios related to resource competition. Meanwhile, most existing microservice approaches address interference either by detecting and localizing performance issues or by optimizing latency alone, without explaining why specific co-locations cause cross-application interference, and how this can inform service placement optimization. This work closes that gap by building a spatio-temporal data structure that captures the causal effects of cross-application interference. These causal effects are mathematically formalized as necessary and sufficient conditional probabilities that inform a multi-objective optimizer (Optuna). Cross-application profiling is used to simulate traces and estimate interference probabilities, while per-service latency baselines are provided by performance data, such as 95th-percentile response times (p95). Our approach supports network penalties, application isolation requirements, and adjustable weighting of necessary and sufficient causal metrics. Experimental results on real multi-application workloads show that interference-aware placements significantly reduce cross-application interference and improve response performance. Ultimately, the causality-driven multi-objective formulation gives cloud operators explicit control over interference, latency, and communication overhead when configuring service placements.

cs.DC↗

arXiv2026

CrossMaps: Confidence-Aware Open-Vocabulary Semantic Mapping for Rover Navigation

Rovers rely on perception to maintain spatial maps that encode both objects and sensor quality (e.g., range reliability, lighting artifacts, data density), guiding data fusion, embedding updates, and navigation under partial observability. To study these coupled perception-navigation processes, we present CrossMaps, a real-time confidence-aware open-vocabulary semantic mapping pipeline that constructs language-queryable maps from RGB-D data. Building on VLMaps-style approaches, CrossMaps integrates multi-scale CLIP embeddings with confidence-aware fusion and a dual-memory architecture consisting of Short-Term Memory (STM) and Long-Term Memory (LTM). The STM aggregates noisy visual observations using geometric, semantic, and temporal confidence cues, while confident and coherent cells are promoted to the LTM as persistent semantic landmarks. Designed for deployment with a Jetson Orin-powered UGV alongside SLAM, CrossMaps runs in real time and produces semantic heatmaps that can be queried with natural language to guide rover navigation.

cs.RO↗

arXiv2026

From Observation to Intervention: A Causal Audit of Expert Importance in Mixture-of-Experts Models

Interpretability methods routinely use population-level summary statistics over observed model behaviour to license claims about the effects of targeted interventions on specific computations; in Pearl's terms, they treat rung-1 associational evidence as if it supported rung-2 interventional conclusions, a move whose validity is rarely tested. We examine one concrete instance: the use of routing statistics in Mixture-of-Experts (MoE) pruning, where utilization rates, activation norms, and routing weight distributions are treated as predictors of which experts can be removed without functional cost. A token-level interventional audit across three high-redundancy MoE architectures (OLMoE-1B-7B-0924, Qwen1.5-MoE-A2.7B, DeepSeek-V2-Lite) finds no observational metric predicts causal expert importance in any model: across all 60 metric-layer combinations effect sizes stay below Cohen's $d = 0.23$, and no metric is reliably positive under our corrected, dual-test criterion. A per-token routing weight control, run with identical $n$, rules out insufficient power, recovering a signal whose CI excludes zero at OLMoE's final MoE layer ($d = +0.231$, 95\% CI $[+0.09, +0.37]$, $p = 0.0013$). Existing pruning methods succeed in this regime not by identifying dispensable experts but because early-layer redundancy renders most selection criteria interchangeable. Our results provide an explicit counterexample to the common inferential step from population-level observational summaries to token-level interventional claims about expert importance, and illustrate how interventional audits can calibrate the evidential standards for interpretability claims.

cs.LG↗

arXiv2026

Towards Neuro-symbolic Causal Rule Synthesis, Verification, and Evaluation Grounded in Legal and Safety Principles

Rule-based systems remain central in safety-critical domains but often struggle with scalability, brittleness, and goal misspecification. These limitations can lead to reward hacking and failures in formal verification, as AI systems tend to optimize for narrow objectives. In previous research, we developed a neuro-symbolic causal framework that integrates first-order logic abduction trees, structural causal models, and deep reinforcement learning within a MAPE-K loop to provide explainable adaptations under distribution shifts. In this paper, we extend that framework by introducing a meta-level layer designed to mitigate goal misspecification and support scalable rule maintenance. This layer consists of a Goal/Rule Synthesizer and a Rule Verification Engine, which iteratively refine a formal rule theory from high-level natural-language goals and principles provided by human experts. The synthesis pipeline employs large language models (LLMs) to: (1) decompose goals into candidate causes, (2) consolidate semantics to remove redundancies, (3) translate them into candidate first-order rules, and (4) compose necessary and sufficient causal sets. The verification pipeline then performs (1) syntax and schema validation, (2) logical consistency analysis, and (3) safety and invariant checks before integrating verified rules into the knowledge base. We evaluated our approach with a proof-of-concept implementation in two autonomous driving scenarios. Results indicate that, given human-specified goals and principles, the pipeline can successfully derive minimal necessary and sufficient rule sets and formalize them as logical constraints. These findings suggest that the pipeline supports incremental, modular, and traceable rule synthesis grounded in established legal and safety principles.

cs.LO↗

arXiv2026

From Program Slices to Causal Clarity: Evaluating Faithful, Actionable LLM-Generated Failure Explanations via Context Partitioning and LLM-as-a-Judge

Large language model (LLM)-based debugging systems can generate failure explanations, but these explanations may be incomplete or incorrect. Misleading explanations are harmful for downstream tasks (e.g., bug triage, bug fixing). We investigate how explanation quality is affected by various LLM context configurations. Existing work predominantly treats LLM-generated failure explanations as an ad hoc by-product of debugging or repair workflows, using generic prompting over undifferentiated artifacts such as code, tests, and error messages rather than targeting explanations as a first-class output with dedicated quality assessment. Consequently, existing approaches provide limited support for assessing whether these explanations capture the underlying fault-error-failure mechanism and for actionable next steps, and most techniques instead prioritize task success (e.g., patch correctness or review quality) over the explicit causal explanation quality. We systematically vary the debugging information to study how distinct context compositions affect the quality of LLM-generated failure explanations. Across 93 context configurations on real bugs and three economically viable models (gpt-5-mini, DeepSeek-V3.2, and Grok-4.1-fast), we evaluate explanations with six criteria and validate the LLM-as-a-judge scores against human ratings in a user study. Our results indicate that explanation quality is causally affected by context composition. Evidence-rich, failure-specific artifacts improve causal and action-oriented quality, whereas overly large contexts tend to yield vague explanations. Higher explanation-score quartiles are associated with higher downstream repair pass rates and, for some models, with fixes that are closer to the reference minimal fixes. In contrast, low-score quartiles can even underperform the no-explanation baseline. Reproduction package is publicly available.

cs.SE↗

arXiv2025

Causal Knowledge Transfer for Multi-Agent Reinforcement Learning in Dynamic Environments

[Context] Multi-agent reinforcement learning (MARL) has achieved notable success in environments where agents must learn coordinated behaviors. However, transferring knowledge across agents remains challenging in non-stationary environments with changing goals. [Problem] Traditional knowledge transfer methods in MARL struggle to generalize, and agents often require costly retraining to adapt. [Approach] This paper introduces a causal knowledge transfer framework that enables RL agents to learn and share compact causal representations of paths within a non-stationary environment. As the environment changes (new obstacles), agents' collisions require adaptive recovery strategies. We model each collision as a causal intervention instantiated as a sequence of recovery actions (a macro) whose effect corresponds to a causal knowledge of how to circumvent the obstacle while increasing the chances of achieving the agent's goal (maximizing cumulative reward). This recovery action macro is transferred online from a second agent and is applied in a zero-shot fashion, i.e., without retraining, just by querying a lookup model with local context information (collisions). [Results] Our findings reveal two key insights: (1) agents with heterogeneous goals were able to bridge about half of the gap between random exploration and a fully retrained policy when adapting to new environments, and (2) the impact of causal knowledge transfer depends on the interplay between environment complexity and agents' heterogeneous goals.

cs.AI↗

arXiv2024

Localized RETE for Incremental Graph Queries with Nested Graph Conditions

The growing size of graph-based modeling artifacts in model-driven engineering calls for techniques that enable efficient execution of graph queries. Incremental approaches based on the RETE algorithm provide an adequate solution in many scenarios, but are generally designed to search for query results over the entire graph. However, in certain situations, a user may only be interested in query results for a subgraph, for instance when a developer is working on a large model of which only a part is loaded into their workspace. In this case, the global execution semantics can result in significant computational overhead. To mitigate the outlined shortcoming, in this article we propose an extension of the RETE approach that enables local, yet fully incremental execution of graph queries, while still guaranteeing completeness of results with respect to the relevant subgraph. We empirically evaluate the presented approach via experiments inspired by a scenario from software development and with queries and data from an independent social network benchmark. The experimental results indicate that the proposed technique can significantly improve performance regarding memory consumption and execution time in favorable cases, but may incur a noticeable overhead in unfavorable cases.

cs.LO↗

arXiv2024

Localized RETE for Incremental Graph Queries

Context: The growing size of graph-based modeling artifacts in model-driven engineering calls for techniques that enable efficient execution of graph queries. Incremental approaches based on the RETE algorithm provide an adequate solution in many scenarios, but are generally designed to search for query results over the entire graph. However, in certain situations, a user may only be interested in query results for a subgraph, for instance when a developer is working on a large model of which only a part is loaded into their workspace. In this case, the global execution semantics can result in significant computational overhead. Contribution: To mitigate the outlined shortcoming, in this paper we propose an extension of the RETE approach that enables local, yet fully incremental execution of graph queries, while still guaranteeing completeness of results with respect to the relevant subgraph. Results: We empirically evaluate the presented approach via experiments inspired by a scenario from software development and an independent social network benchmark. The experimental results indicate that the proposed technique can significantly improve performance regarding memory consumption and execution time in favorable cases, but may incur a noticeable linear overhead in unfavorable cases.

cs.SE↗

arXiv2024

Mind the Gap Between Synthetic and Real: Utilizing Transfer Learning to Probe the Boundaries of Stable Diffusion Generated Data

Generative foundation models like Stable Diffusion comprise a diverse spectrum of knowledge in computer vision with the potential for transfer learning, e.g., via generating data to train student models for downstream tasks. This could circumvent the necessity of collecting labeled real-world data, thereby presenting a form of data-free knowledge distillation. However, the resultant student models show a significant drop in accuracy compared to models trained on real data. We investigate possible causes for this drop and focus on the role of the different layers of the student model. By training these layers using either real or synthetic data, we reveal that the drop mainly stems from the model's final layers. Further, we briefly investigate other factors, such as differences in data-normalization between synthetic and real, the impact of data augmentations, texture vs.\ shape learning, and assuming oracle prompts. While we find that some of those factors can have an impact, they are not sufficient to close the gap towards real data. Building upon our insights that mainly later layers are responsible for the drop, we investigate the data-efficiency of fine-tuning a synthetically trained model with real data applied to only those last layers. Our results suggest an improved trade-off between the amount of real training data used and the model's accuracy. Our findings contribute to the understanding of the gap between synthetic and real data and indicate solutions to mitigate the scarcity of labeled real data.

cs.CV↗

arXiv2023

Incremental Model Transformations with Triple Graph Grammars for Multi-version Models

Like conventional software projects, projects in model-driven software engineering require adequate management of multiple versions of development artifacts, importantly allowing living with temporary inconsistencies. In previous work, multi-version models for model-driven software engineering have been introduced, which allow checking well-formedness and finding merge conflicts for multiple versions of a model at once. However, also for multi-version models, situations where different artifacts, that is, different models, are linked via automatic model transformations have to be handled. In this paper, we propose a technique for jointly handling the transformation of multiple versions of a source model into corresponding versions of a target model, which enables the use of a more compact representation that may afford improved execution time of both the transformation and further analysis operations. Our approach is based on the well-known formalism of triple graph grammars and the aforementioned encoding of model version histories called multi-version models. In addition to batch transformation of an entire model version history, the technique also covers incremental synchronization of changes in the framework of multi-version models. We show the correctness of our approach with respect to the standard semantics of triple graph grammars and conduct an empirical evaluation to investigate the performance of our technique regarding execution time and memory consumption. Our results indicate that the proposed technique affords lower memory consumption and may improve execution time for batch transformation of large version histories, but can also come with computational overhead in unfavorable cases.

cs.SE↗

arXiv2023

Triple Graph Grammars for Multi-version Models

Like conventional software projects, projects in model-driven software engineering require adequate management of multiple versions of development artifacts, importantly allowing living with temporary inconsistencies. In the case of model-driven software engineering, employed versioning approaches also have to handle situations where different artifacts, that is, different models, are linked via automatic model transformations. In this report, we propose a technique for jointly handling the transformation of multiple versions of a source model into corresponding versions of a target model, which enables the use of a more compact representation that may afford improved execution time of both the transformation and further analysis operations. Our approach is based on the well-known formalism of triple graph grammars and a previously introduced encoding of model version histories called multi-version models. In addition to showing the correctness of our approach with respect to the standard semantics of triple graph grammars, we conduct an empirical evaluation that demonstrates the potential benefit regarding execution time performance.

cs.SE↗

arXiv2023

Modular and Incremental Global Model Management with Extended Generalized Discrimination Networks

Complex projects developed under the paradigm of model-driven engineering nowadays often involve several interrelated models, which are automatically processed via a multitude of model operations. Modular and incremental construction and execution of such networks of models and model operations are required to accommodate efficient development with potentially large-scale models. The underlying problem is also called Global Model Management. In this report, we propose an approach to modular and incremental Global Model Management via an extension to the existing technique of Generalized Discrimination Networks (GDNs). In addition to further generalizing the notion of query operations employed in GDNs, we adapt the previously query-only mechanism to operations with side effects to integrate model transformation and model synchronization. We provide incremental algorithms for the execution of the resulting extended Generalized Discrimination Networks (eGDNs), as well as a prototypical implementation for a number of example eGDN operations. Based on this prototypical implementation, we experiment with an application scenario from the software development domain to empirically evaluate our approach with respect to scalability and conceptually demonstrate its applicability in a typical scenario. Initial results confirm that the presented approach can indeed be employed to realize efficient Global Model Management in the considered scenario.

cs.SE↗

arXiv2022

Towards Development with Multi-Version Models: Detecting Merge Conflicts and Checking Well-Formedness

Developing complex software requires that multiple views and versions of the software can be developed in parallel and merged as supported by views and managed by version control systems. In this context, this paper considers monitoring merging and related consistency problems permanently at the level of models and abstract syntax to permit early and frequent conflict detection while developing in parallel. The presented approach introduces multi-version models based on typed graphs that permit to store changes and multiple versions in one graph in a compact form and allow (1) to study well-formedness for all versions without the need to extract each version individually, (2) to report all possible merge conflicts without the need to merge all pairs of versions, and (3) to report all violations of well-formedness conditions that will result for merges of any two versions independent of any merge decisions without the need to merge all pairs of versions. The paper defines the related concepts and algorithms operating on multi-version models, proves their correctness w.r.t.~the usually employed three-way-merge, and reports on preliminary experiments concerning the scalability.

cs.SE↗

arXiv2021

Hybrid Planning with Receding Horizon: A Case for Meta-self-awareness

The trade-off between the quality and timeliness of adaptation is a multi-faceted challenge in engineering self-adaptive systems. Obtaining adaptation plans that fulfill system objectives with high utility and in a timely manner is the holy grail, however, as recent research revealed, it is not trivial. Hybrid planning is concerned with resolving the time and quality trade-off via dynamically combining multiple planners that individually aim to perform either timely or with high quality. The choice of the most fitting planner is steered based on assessments of runtime information. A hybrid planner for a self-adaptive system requires (i) a decision-making mechanism that utilizes (ii) system-level as well as (iii) feedback control-level information at runtime. In this paper, we present HYPEZON, a hybrid planner for self-adaptive systems. Inspired by model predictive control, HYPEZON leverages receding horizon control to utilize runtime information during its decision-making. Moreover, we propose to engineer HYPEZON for self-adaptive systems via two alternative designs that conform to meta-self-aware architectures. Meta-self-awareness allows for obtaining knowledge and reasoning about own awareness via adding a higher-level reasoning entity. HYPEZON aims to address the problem of hybrid planning by considering it as a case for meta-self-awareness.

cs.SE↗

arXiv2021

Probabilistic Metric Temporal Graph Logic

Cyber-physical systems often encompass complex concurrent behavior with timing constraints and probabilistic failures on demand. The analysis whether such systems with probabilistic timed behavior ad-here to a given specification is essential. When the states of the system can be represented by graphs, the rule-based formalism of Probabilistic Timed Graph Transformation Systems (PTGTSs) can be used to suitably capture structure dynamics as well as probabilistic and timed behavior of the system. The model checking support for PTGTSs w.r.t. properties specified using Probabilistic Timed Computation Tree Logic (PTCTL) has been already presented. Moreover, for timed graph-based runtime monitoring, Metric Temporal Graph Logic (MTGL) has been developed for stating metric temporal properties on identified subgraphs and their structural changes over time. In this paper, we (a) extend MTGL to the Probabilistic Metric Temporal Graph Logic (PMTGL) by allowing for the specification of probabilistic properties, (b) adapt our MTGL satisfaction checking approach to PTGTSs, and (c) combine the approaches for PTCTL model checking and MTGL satisfaction checking to obtain a Bounded Model Checking (BMC) approach for PMTGL. In our evaluation, we apply an implementation of our BMC approach in AutoGraph to a running example.

cs.SE↗

arXiv2020

A Scalable Querying Scheme for Memory-efficient Runtime Models with History

Runtime models provide a snapshot of a system at runtime at a desired level of abstraction. Via a causal connection to the modeled system and by employing model-driven engineering techniques, runtime models support schemes for (runtime) adaptation where data from previous snapshots facilitates more informed decisions. Nevertheless, although runtime models and model-based adaptation techniques have been the focus of extensive research, schemes that treat the evolution of the model over time as a first-class citizen have only lately received attention. Consequently, there is a lack of sophisticated technology for such runtime models with history. We present a querying scheme where the integration of temporal requirements with incremental model queries enables scalable querying for runtime models with history. Moreover, our scheme provides for a memory-efficient storage of such models. By integrating these two features into an adaptation loop, we enable efficient history-aware self-adaptation via runtime models, of which we present an implementation.

cs.SE↗

arXiv2020

Improving Scalability and Reward of Utility-Driven Self-Healing for Large Dynamic Architectures

Self-adaptation can be realized in various ways. Rule-based approaches prescribe the adaptation to be executed if the system or environment satisfies certain conditions. They result in scalable solutions but often with merely satisfying adaptation decisions. In contrast, utility-driven approaches determine optimal decisions by using an often costly optimization, which typically does not scale for large problems. We propose a rule-based and utility-driven adaptation scheme that achieves the benefits of both directions such that the adaptation decisions are optimal, whereas the computation scales by avoiding an expensive optimization. We use this adaptation scheme for architecture-based self-healing of large software systems. For this purpose, we define the utility for large dynamic architectures of such systems based on patterns that define issues the self-healing must address. Moreover, we use pattern-based adaptation rules to resolve these issues. Using a pattern-based scheme to define the utility and adaptation rules allows us to compute the impact of each rule application on the overall utility and to realize an incremental and efficient utility-driven self-healing. In addition to formally analyzing the computational effort and optimality of the proposed scheme, we thoroughly demonstrate its scalability and optimality in terms of reward in comparative experiments with a static rule-based approach as a baseline and a utility-driven approach using a constraint solver. These experiments are based on different failure profiles derived from real-world failure logs. We also investigate the impact of different failure profile characteristics on the scalability and reward to evaluate the robustness of the different approaches.

cs.SE↗

arXiv2020

Collective Risk Minimization via a Bayesian Model for Statistical Software Testing

In the last four years, the number of distinct autonomous vehicles platforms deployed in the streets of California increased 6-fold, while the reported accidents increased 12-fold. This can become a trend with no signs of subsiding as it is fueled by a constant stream of innovations in hardware sensors and machine learning software. Meanwhile, if we expect the public and regulators to trust the autonomous vehicle platforms, we need to find better ways to solve the problem of adding technological complexity without increasing the risk of accidents. We studied this problem from the perspective of reliability engineering in which a given risk of an accident has severity and probability of occurring. Timely information on accidents is important for engineers to anticipate and reuse previous failures to approximate the risk of accidents in a new city. However, this is challenging in the context of autonomous vehicles because of the sparse nature of data on the operational scenarios (driving trajectories in a new city). Our approach was to mitigate data sparsity by reducing the state space through monitoring of multiple-vehicles operations. We then minimized the risk of accidents by determining proper allocation of tests for each equivalence class. Our contributions comprise (1) a set of strategies to monitor the operational data of multiple autonomous vehicles, (2) a Bayesian model that estimates changes in the risk of accidents, and (3) a feedback control-loop that minimizes these risks by reallocating test effort. Our results are promising in the sense that we were able to measure and control risk for a diversity of changes in the operational scenarios. We evaluated our models with data from two real cities with distinct traffic patterns and made the data available for the community.

cs.SE↗