process tracing

To address the challenges of inconsistent modeling of alternating causal conditions in multi-variant event logs, insufficient robustness to missing values, and ambiguous cross-trace causal logic representation, this paper proposes the first formally verifiable causal process model fusion method. Our approach achieves lossless unification of multiple causal process variants for the first time, explicitly encoding alternating causal flows among activity traces. It integrates event log segmentation, missing-value-robust preprocessing, and graph-structured model fusion—grounded in causal discovery, process mining, and temporal logic modeling. Evaluation on three public and two private datasets demonstrates significant improvements in consistency of cross-variant causal explanations and feasibility of business-level interventions. The open-source implementation ensures full reproducibility.

Industrial research agents often generate experimental trajectories containing invalid or incomplete information, rendering them unreliable for direct decision-making. This work proposes an evidence-oriented framework that automatically transforms such trajectories into structured evidence through a context-isolated generate–verify–repair pipeline. The approach introduces intervention-level claim categorization—distinguishing actionable repairs, diagnostic safeguards, and retained discoveries—and incorporates end-to-end provenance tracking to enable claim scoping and auditability. Experimental results demonstrate that the resulting candidate solutions outperform existing baselines. Audits further reveal that trajectory evolution is non-monotonic, and that applicability assessment constitutes a key performance bottleneck for the controller.

Electronic health record (EHR) audit logs contain rich, multidimensional information about clinical activities, yet lack a unified modeling framework. This work proposes a “multi-axis trace” perspective that simultaneously associates each logged action with clinician behavior, patient care trajectories, team collaboration patterns, and repetitive workflow structures, thereby uncovering its multifaceted clinical semantics. Building on this insight, we develop a representation learning framework that preserves the multi-axis structure by pretraining a foundation model directly on raw audit log streams to learn general-purpose representations. The resulting approach establishes a unified data representation and evaluation paradigm applicable to diverse downstream tasks, including clinical workload analysis, patient outcome prediction, team coordination assessment, and workflow modeling.

Causal Loop Diagram (CLD) construction in system dynamics suffers from low efficiency and high entry barriers for novices. Method: This paper proposes the first stepwise prompt engineering framework tailored for CLD generation, leveraging large language models (LLMs) to automatically map textual dynamic hypotheses into structured CLDs. The approach integrates chain-of-thought reasoning, role-guided prompting, and domain-specific constraints, representing CLDs as standard directed graphs; it is fine-tuned and evaluated on a textbook-based system dynamics dataset. Contribution/Results: Experiments show that the automatically generated CLDs achieve 89% agreement with expert-built diagrams on simple dynamic structures, substantially reducing modeling time. This work establishes the first end-to-end, accurate, interpretable, and domain-aligned natural-language-to-CLD generation pipeline, empirically validating the feasibility and practical utility of LLMs in automating system modeling.

This study addresses the challenges posed by fragmented and heterogeneous data in child sexual exploitation and abuse (CSEA) cases, which hinder cross-case analysis and impose significant emotional burdens on investigators. To overcome these issues, the authors propose a modular open-source system that combines regular expressions with semantic pattern analysis to enable interpretable and auditable extraction of structured information, thereby constructing a unified case data model. The system employs a multidimensional weighted Jaccard similarity measure for case clustering and integrates six types of interactive visualizations—including timelines and severity indicators—to support automated triage and in-depth investigative reasoning. Evaluation on 47 AZICAC reports (2011–2014) demonstrates that the system substantially improves cross-case analytical efficiency and effectively reduces the manual handling of sensitive content.

This work addresses the challenge of providing trustworthy explanations for deep learning models in predictive process monitoring, where their black-box nature and the limitations of existing attribution methods hinder both computational efficiency for long traces and semantic fidelity to control-flow dynamics. The authors propose a novel local post-hoc interpretability approach that leverages control-flow structures to semantically segment event logs and computes SHAP attributions over these segments to identify critical process fragments and turning points influencing predictions. Experimental results demonstrate that the method accurately captures known logical turning points on synthetic datasets and effectively uncovers the underlying dynamic mechanisms driving predictions in real-world loan approval and municipal process logs, achieving a balanced trade-off between computational efficiency and process-aware semantic interpretability.

Existing process mining approaches struggle to capture cross-case decision synchronization mechanisms, which are critical for equitable resource allocation and process efficiency. This work addresses this gap by formally defining and implementing an automated method for discovering four distinct types of decision synchronization patterns—thereby filling a key void in traditional process mining that overlooks inter-case dependencies. Drawing inspiration from supply chain coordination, the authors propose a unified framework that integrates event log analysis, process structure modeling, and execution constraint reasoning to uncover such synchronization behaviors. Experimental evaluation on two synthetic scenarios demonstrates that the proposed method accurately reconstructs all four synchronization patterns, confirming its effectiveness and scalability.

This study addresses the lack of systematic preprocessing standards, integrated analytical workflows, and cross-method consistency checks in current computer-based assessment process data. To bridge this gap, the authors propose an end-to-end analytical framework featuring a unified preprocessing pipeline and a dual-path analysis paradigm that synergistically combines feature engineering with model-based inference. The framework incorporates large language models (LLMs) to standardize action sequences and facilitate process-data-driven differential item functioning (DIF) detection. Technically, it integrates timestamp correction, action chunking, n-gram and TF-IDF feature extraction, multidimensional scaling, hidden Markov modeling, and subtask identification. Empirical results demonstrate that n-gram–based behavioral clustering offers diagnostic value for incorrect responders, multidimensional scaling effectively reconstructs behavioral constructs, and process data can identify and mitigate construct-irrelevant group differences.

This work addresses the limited interpretability of large language model–driven automated program repair, which hinders diagnosis and reproducibility of failure cases. To this end, we present TraceView, the first interactive visualization tool that structures repair trajectories into Thought-Action-Result triplets and supports semantic relationship annotation. By integrating trajectory parsing, relational modeling, and graph-based visualization techniques, TraceView enables traceable analysis from high-level overviews to fine-grained details. A user study demonstrates that TraceView significantly enhances developers’ comprehension of the repair process and improves navigation efficiency. The implementation and a demonstration video are publicly available.