Pith. sign in

REVIEW 4 major objections 4 minor 3 cited by

From Theory to Practice: Real-World Use Cases on Trustworthy LLM-Driven Process Modeling, Prediction and Automation

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper argues that LLMs, kept under human supervision and grounded in process data, can support every phase of the business process management lifecycle, with four industrial use cases as evidence.

desk verdict Four early-stage LLM-for-BPM use cases with honest limitations, but the abstract and conclusion claim more than the evidence supports. read the letter →

arxiv 2506.03801 v1 pith:FY6KMEV3 submitted 2025-06-04 cs.SE cs.LGcs.MA

Peter Pfeiffer , Alexander Rombach , Maxim Majlatow , Nijat Mehdiyev This is my paper · ORCID
classification cs.SEcs.LGcs.MA
keywords BusinessProcessManagementPredictionModelingLargeLanguageModelsExplainableAIHumanComputerInteractionKnowledgeGraphsPharmacovigilance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that large language models, embedded in human-in-the-loop workflows and grounded in verifiable data, can support every phase of the business process management lifecycle, and it backs this claim with four industrial pilot use cases. The cases span manufacturing process-prediction dialogue, conversational BPMN process modeling, knowledge-graph-augmented pharmacovigilance agents, and a multi-agent system for sustainable textile design. For each, the paper describes an architecture that combines LLMs with process mining, uncertainty quantification, retrieval-augmented generation, or knowledge graphs, and reports early experiences rather than mature evaluations. The intended contribution is practical and architectural: a context-sensitive integration recipe, not a universal solution. If the claim holds, LLM-based assistants could lower modeling barriers, make predictive process monitoring auditable, and relieve regulatory experts of repetitive screening while keeping a human in the loop.

What carries the argument

The central mechanism is a human-in-the-loop LLM agent architecture layered over process-aware data. Across the four use cases, the load-bearing components are: process mining and event logs providing temporal context; uncertainty quantification and explainability methods such as Shapley values, conformal prediction, and counterfactuals attached to ML predictions; retrieval-augmented generation and knowledge graphs anchoring LLM outputs in verifiable sources; and orchestrators that compose specialized agents under a central controller. This combination converts a single LLM call into a repeatable, inspectable workflow whose outputs can be validated before a human accepts them.

What would settle it

Run the proposed architectures on a shared benchmark: have the conversational BPMN system generate models from one hundred process descriptions and the pharmacovigilance agent answer a set of drug-safety queries, then count XML-valid models, missing sequence flows, and answer accuracy against expert-curated ground truth; if even the larger LLMs fail a large fraction of these checks, the claim that LLMs can support all BPM lifecycle phases loses its empirical support.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the BPM lifecycle need no longer be divided into a discovery phase for human experts and an execution phase for software; the same conversational, agentic LLM infrastructure can address all phases. It reports that process predictions can be turned into auditable dialogues by coupling uncertainty-aware machine learning with process-mining context and retrieval-augmented explanations; that chat interfaces can generate BPMN models but currently produce invalid XML and missing paths; that LLM agents can formulate knowledge-graph queries for drug-safety data; and that multi-agent orchestration can surface sustainability trade-offs for textile designers. The authors conclude that with trust mechanisms and mandatory human supervision, all phases of the lifecycle can be supported.

Load-bearing premise

The load-bearing premise is that LLM outputs can be made sufficiently reliable and auditable through grounding techniques such as retrieval-augmented generation, knowledge graphs, and uncertainty quantification, plus human review, a premise that the paper's own pilots show is still being tested rather than proven.

Editorial extensions

If this is right

  • LLM-based chat interfaces could let domain experts without BPMN training create, query, and refine process models, reducing the modeling bottleneck that currently sits with notation specialists.
  • Process predictions in manufacturing could be delivered with confidence intervals and natural-language explanations that planners can interrogate, which should increase the adoption of predictive process monitoring.
  • Knowledge-graph-grounded agents could take over large parts of routine pharmacovigilance screening, leaving only ambiguous cases for human regulators.
  • Multi-agent systems could turn sustainability regulation into design-time recommendations, making trade-offs such as durability versus recyclability explicit before production decisions are made.
  • The same trust mechanisms, retrieval-augmented generation, uncertainty quantification, and human review, would need to be standard components of any LLM-based BPM product rather than optional add-ons.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's own evidence suggests the 'all phases' claim is a research agenda rather than a demonstrated result; a fair reading is that the load-bearing work lies in showing where LLMs fail, such as BPMN validity, and where grounding mechanisms could fill the gap.
  • A testable extension would be a benchmark where one system must discover, model, execute, monitor, and redesign a single real process while measuring user trust and output validity at each phase; no current use case covers the complete loop.
  • The four use cases point to a common failure mode, namely that LLM-generated artifacts such as XML models and agent queries need a validator or refinement loop before reaching users; this could be generalized into a verification layer for process-aware LLMs.
  • If the human-in-the-loop requirement holds across industries, the economic value of LLMs in BPM may depend less on raw model size and more on the quality of the surrounding grounding and validation infrastructure.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper presents four use cases for applying large language models (LLMs) to business process management (BPM): an LLM-mediated interaction framework for trustworthy process predictions in manufacturing, a conversational BPMN modeling system, a knowledge-graph-augmented LLM agent system for pharmacovigilance, and a multi-agent assistant for sustainable textile design. For each use case, the paper describes the motivation, proposed architecture, and reported experiences, and it includes a comparative discussion of challenges, trust mechanisms, and future directions. The stated central claim, in the introduction and conclusion, is that these use cases demonstrate that all phases of the BPM lifecycle can be supported by LLMs with human-in-the-loop and explainable-AI mechanisms.

Significance. If the demonstration claim were fully supported, the paper would be a useful practical mapping of LLM architectures to BPM lifecycle phases across manufacturing, consulting, life-science, and design domains, with a clear emphasis on trustworthiness and human oversight. The paper's strengths are its coverage of four diverse industrial settings, its explicit acknowledgment of limitations including invalid BPMN XML, hallucinated tags, and unvalidated agent workflows, and its comparison table that organizes input modalities, trust mechanisms, and challenges. However, the evidence presented is preliminary: two use cases explicitly defer validation, and the other two reference prior papers instead of reporting concrete metrics. The significance is therefore conditional on whether the claims are reframed as an early-stage research agenda or supplemented with evaluation results.

major comments (4)
  1. [1 and 7] The paper's central claim, stated in Section 1 and restated in Section 7, is that the four use cases 'demonstrate' that all BPM lifecycle phases can be supported. The evidence reported in Sections 4.3 and 5.3 explicitly defers validation: Section 4.3 says 'Next steps include validating the quality of the generated answers' and Section 5.3 says 'Whether LLMs can balance these trade-offs remains to be seen.' With two of four use cases lacking outcome evaluations and the other two offering only qualitative experience reports, the word 'demonstrate' overstates what the manuscript establishes. Please either add concrete evaluation results for each use case or reframe Sections 1, 6, and 7 as proposing early-stage use cases and a research agenda, rather than demonstrations.
  2. [3.3] The conversational BPMN use case reports that 'LLMs struggle with BPMN-XML' and that generation 'results in many invalid models' and includes hallucinated XML tags, yet the paper provides no quantitative validity rates, semantic accuracy scores, or user study results; the details are deferred to reference [18]. If this use case is to support the 'demonstrate' claim, the paper needs at least the reported validity rates, model quality metrics, or a comparison of model sizes, or it must be explicitly characterized as a lessons-learned pilot with the central claim narrowed accordingly.
  3. [2.3] The manufacturing use case's experience paragraph says 'Early experiments suggest' improvements in trust and interpretability, but no experimental setup or measured outcomes appear in the paper; the claims rest on self-citations [24-28]. Please include the actual evidence, such as number of participants, prediction accuracy, uncertainty quantification metrics, or interaction log analyses, or state clearly that these are observations from prior work rather than results established in this paper.
  4. [4.3] The pharmacovigilance use case is described as feasible, but the manuscript states that answer quality and agent workflow performance still need validation. The claim that this use case demonstrates support for the monitoring phase of the BPM lifecycle is therefore premature. Please either report an evaluation of KG-query accuracy and agent outputs or adjust the phase-coverage claim to 'planned' rather than 'demonstrated.'
minor comments (4)
  1. [Abstract and 1] The abstract and Section 1 use 'demonstrate' in the present tense, while the abstract also says 'We intend to examine tensions'; this tense mismatch between intended work and claimed demonstration should be resolved to avoid overstating the contributions.
  2. [Section 7] The phrase 'In all cases, it emerged that human-machine interactions are essential' is vague; please specify how this conclusion was observed or measured across the four use cases.
  3. [Table 1 and Figure 1] In Figure 1, the label 'UQ-ModelsML-Models' appears to be a typo and should likely read 'UQ-Models / ML-Models'; also, Table 1 lists lifecycle phases that do not match the 'discovery and execution to monitoring and optimization' phrasing in Section 1, so the lifecycle terminology should be harmonized.
  4. [References] Reference [14] lists 'Yelong shen' with a lowercase surname and inconsistent capitalization; in addition, the ACM Reference Format line contains 'InProceedings' without a space, and several references use inconsistent spelling such as 'Process Modelling' versus 'process modeling.'

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a qualitative use-case survey; its self-citations point to separate empirical studies, while the cited limitations are evidentiary gaps, not constructed predictions.

full rationale

The paper contains no derivations, equations, fitted parameters, or numerical predictions, so there is no equation-level reduction of a claimed result to its inputs. The central claim—'demonstrating that all phases of the BPM lifecycle can be supported' (Section 7)—is supported by qualitative project descriptions and by citations to the authors' earlier published work: Section 2.3 says 'Early experiments suggest that combining robust data pipelines with iterative user interfaces significantly enhances trust among domain experts [24–28]', and Section 3.3 reports findings 'see [18] for more details'. These citations reference separately published empirical studies, not results defined in terms of the present paper's conclusion, so they constitute external evidence rather than circular self-justification. The manuscript itself flags the load-bearing evidential weaknesses: Section 3.3 admits LLMs 'struggle with BPMN-XML' and 'results in many invalid models'; Section 4.3 lists 'validating the quality of the generated answers' as a next step; Section 5.3 says whether LLMs can balance sustainability trade-offs 'remains to be seen'; and Section 2.3 offers only 'early experiments suggest'. These are limitations in demonstrated feasibility, not circular reasoning. No uniqueness theorem, ansatz, or fitted value is imported from the authors' prior work to force the paper's claims. Under the hard-evidence rule, no circular step can be exhibited; the honest finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper contains no derivations, so there are no fitted parameters. The load-bearing premises are domain assumptions about LLM reliability, grounding, and human oversight, none of which are tested with independent data in this paper. No new theoretical entities are introduced.

assumptions (3)
  • domain assumption LLM outputs can be grounded and made auditable through retrieval-augmented generation, uncertainty quantification, and knowledge graph integration.
    Relied on in all four use cases as the mechanism for trustworthiness; no evaluation data is provided in this paper to show the grounding works.
  • domain assumption Human-in-the-loop supervision will catch errors made by LLM agents, such as invalid BPMN XML or hallucinations.
    Section 6.1 states 'human interaction is enforced' and 'no decisions are made without human supervision,' but the paper does not test whether users can detect the errors that its own experiments report.
  • domain assumption The four early-stage projects are representative of the BPM lifecycle phases they are mapped to, supporting generalization to 'all phases of the BPM lifecycle.'
    Section 7 generalizes from four use cases to all BPM lifecycle phases; two of the use cases have no completed results, so representativeness is not established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Theory to Practice: Real-World Use Cases on Trustworthy LLM-Driven Process Modeling, Prediction and Automation." pith.science (2026). https://pith.science/paper/FY6KMEV3

@misc{pith2026250603801,
  author       = {Pith},
  title        = {Pith review of: From Theory to Practice: Real-World Use Cases on Trustworthy LLM-Driven Process Modeling, Prediction and Automation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FY6KMEV3}},
  note         = {Machine review of arXiv:2506.03801}
}
read the original abstract

Traditional Business Process Management (BPM) struggles with rigidity, opacity, and scalability in dynamic environments while emerging Large Language Models (LLMs) present transformative opportunities alongside risks. This paper explores four real-world use cases that demonstrate how LLMs, augmented with trustworthy process intelligence, redefine process modeling, prediction, and automation. Grounded in early-stage research projects with industrial partners, the work spans manufacturing, modeling, life-science, and design processes, addressing domain-specific challenges through human-AI collaboration. In manufacturing, an LLM-driven framework integrates uncertainty-aware explainable Machine Learning (ML) with interactive dialogues, transforming opaque predictions into auditable workflows. For process modeling, conversational interfaces democratize BPMN design. Pharmacovigilance agents automate drug safety monitoring via knowledge-graph-augmented LLMs. Finally, sustainable textile design employs multi-agent systems to navigate regulatory and environmental trade-offs. We intend to examine tensions between transparency and efficiency, generalization and specialization, and human agency versus automation. By mapping these trade-offs, we advocate for context-sensitive integration prioritizing domain needs, stakeholder values, and iterative human-in-the-loop workflows over universal solutions. This work provides actionable insights for researchers and practitioners aiming to operationalize LLMs in critical BPM environments.

Figures

Figures reproduced from arXiv: 2506.03801 by the authors.

Figure 1
Figure 1. Three-module AI solution that integrates human-centered ML, LLM–based interaction, and an intelligent AI interface [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Architecture for Conversational Process Modeling [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Architecture of PV KG and LLM-Agents These tasks require screening large amounts of textual reports and their assessment with domain knowledge. For instance, they in￾volve querying and scanning relevant documents in public databases such as PubMed (scientific reports) and MAUDE (reports from med￾ical experts and patients). Further, internal company case reports or social media posts can also be of interest. Domain k… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Architecture of Multi-Agent Textile Design Assistance System [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Assessing the Business Process Modeling Competences of Large Language Models

    cs.SE 2026-01 conditional novelty 6.0 of 10

    Open-source LLMs can produce BPMN process models that rival human experts on syntax and readability, but they lag on semantic accuracy and frequently generate invalid BPMN-XML.

  2. On LLM-Assisted Generation of Smart Contracts from Business Processes

    cs.SE 2025-07 conditional novelty 6.0 of 10

    An open benchmark shows top LLMs generate functionally correct smart contracts from BPMN process models in about 80 to 92 percent of checks, below the reliability needed for blockchain deployment.

  3. Multi-Modal Vision vs. Text-Based Parsing: Benchmarking LLM Strategies for Invoice Processing

    cs.CL 2025-08 conditional novelty 4.0 of 10

    Across three invoice datasets, multimodal LLMs extract fields more accurately from raw images than from markdown converted by a parsing tool, with Gemini 2.5 Pro leading.

Reference graph

Works this paper leans on

44 extracted references · 28 canonical work pages · cited by 3 Pith papers

  1. [18]

    Chantale Lauer, Peter Pfeiffer, Alexander Rombach, and Nijat Mehdiyev. 2025. Conversational Business Process Modeling using LLMs: Initial Results and Chal- lenges. InEMISA 2025. Gesellschaft für Informatik e.V., Bonn. doi:10.18420/ EMISA2025_03

    work page 2025

  2. [1]

    Stefano Abbate, Piera Centobelli, Roberto Cerchione, Simon Peter Nadeem, and Emanuela Riccio. 2023. Sustainability trends and gaps in the textile, apparel and fashion industries.Environment, Development and Sustainability26, 2 (feb 2023), 2837–2864. doi:10.1007/s10668-022-02887-2

    work page doi:10.1007/s10668-022-02887-2 2023

  3. [2]

    Kamala Aliyeva and Nijat Mehdiyev. 2024. Uncertainty-aware multi-criteria decision analysis for evaluation of explainable artificial intelligence methods: A use case from the healthcare domain.Information sciences657 (2024), 119987. doi:10.1016/j.ins.2023.119987

    work page arXiv 2024

  4. [3]

    EUROPEAN COMMISSION. 2022. EU Strategy for Sustainable and Circular Textiles. https://ec.europa.eu/eurostat

    work page 2022

  5. [4]

    2022.Questions and Answers on EU Strategy for Sustainable and Circular Textiles

    European Commission. 2022.Questions and Answers on EU Strategy for Sustainable and Circular Textiles. Technical Report. EU

    work page 2022

  6. [5]

    Chiara Di Francescomarino, Chiara Ghidini, Fabrizio Maria Maggi, and Fredrik Milani. 2018. Predictive Process Monitoring Methods: Which One Suits Me Best?. InBusiness Process Management, Mathias Weske, Marco Montali, Ingo Weber, and Jan vom Brocke (Eds.). Springer International Publishing, Cham, 462–479. doi:10.1007/978-3-319-98648-7_27

    work page doi:10.1007/978-3-319-98648-7_27 2018

  7. [6]

    Marlon Dumas, Marcello La Rosa, Jan Mendling, and Hajo A. Reijers. 2018. In- troduction to Business Process Management.Fundamentals of Business Process Management(2018), 1–33. doi:10.1007/978-3-662-56509-4_1

    work page doi:10.1007/978-3-662-56509-4_1 2018

  8. [7]

    EMA. 2021. EMA Pharmacovigilance System Manual (Version 1.3). (2021). https: //www.ema.europa.eu/en/human-regulatory-overview/pharmacovigilance- overview

    work page 2021

Show all 44 references
  1. [8]

    Jerome H Friedman. 2001. Greedy function approximation: a gradient boosting machine.Annals of statistics(2001), 1189–1232. doi:10.1214/aos/1013203451

    2001

  2. [9]

    Yarin Gal and Zoubin Ghahramani. 2016. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. InProceedings of the 33rd International Conference on International Conference on Machine Learning - Volume 48(New York, NY, USA)(ICML’16). JMLR.org, 1050–1059

    2016

  3. [10]

    Louie Giray. 2023. Prompt engineering with ChatGPT: a guide for academic writers.Annals of biomedical engineering51, 12 (2023), 2629–2633

    2023

  4. [11]

    Alex Goldstein, Adam Kapelner, Justin Bleich, and Emil Pitkin. 2015. Peeking inside the black box: Visualizing statistical learning with plots of individual conditional expectation.Journal of Computational and Graphical Statistics24, 1 (2015), 44–65. doi:10.1080/10618600.2014.907095

    2015

  5. [12]

    Michael Grohs, Luka Abb, Nourhan Elsayed, and Jana-Rebecca Rehse. 2024. Large Language Models Can Accomplish Business Process Management Tasks. InBusiness Process Management Workshops, Jochen De Weerdt and Luise Pufahl (Eds.). Springer Nature Switzerland, Cham, 453–465. doi:10...

    2024 doi

  6. [13]

    Vikas Hassija, Vinay Chamola, Atmesh Mahapatra, Abhinandan Singal, Divyansh Goel, Kaizhu Huang, Simone Scardapane, Indro Spinelli, Mufti Mahmud, and Amir Hussain. 2024. Interpreting Black-Box Models: A Review on Explainable Artificial Intelligence.Cognitive Computation16, 1 (j...

    2024 doi

  7. [14]

    Edward J Hu, Yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-Rank Adaptation of Large Language Models. InInternational Conference on Learning Representations. https: //openreview.net/forum?id=nZeVKeeFYf9

    2022

  8. [15]

    Nataliia Klievtsova, Janik-Vasily Benzin, Timotheus Kampik, Juergen Mangler, and Stefanie Rinderle-Ma. 2023. Conversational Process Modelling: State of the Art, Applications, and Implications in Practice. InBusiness Process Management Forum, Chiara Di Francescomarino, Andrea B...

    2023

  9. [16]

    Humam Kourani, Alessandro Berti, Daniel Schuster, and Wil M P van der Aalst

  10. [17]

    Aleksandra Kuzior, Mariya Sira, and Paulina Brożek. 2023. Use of Artificial Intelligence in Terms of Open Innovation Process and Management.Sustainability 15, 9 (apr 2023), 7205. doi:10.3390/su15097205

    2023 doi

  11. [19]

    Henrik Leopold. 2013. Generation of Natural Language Texts from Process Models. InNatural Language in Business Process Models. Vol. 168. 103–136. doi:10.1007/978- 3-319-04175-9_5

    2013 doi

  12. [20]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-augmented generation for knowledge-intensive NLP tasks. InProceedings o...

    2020

  13. [21]

    Vera Liao, Daniel Gruen, and Sarah Miller

    Q. Vera Liao, Daniel Gruen, and Sarah Miller. 2020. Questioning the AI: Informing Design Practices for Explainable AI User Experiences. InProceedings of the 2020 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA) (CHI ’20). Association for Computing Machin...

    2020

  14. [22]

    Scott M Lundberg and Su-In Lee. 2017. A unified approach to interpreting model predictions.Advances in neural information processing systems30 (2017)

    2017

  15. [23]

    Fabrizio Maria Maggi, Chiara Di Francescomarino, Marlon Dumas, and Chiara Ghidini. 2014. Predictive Monitoring of Business Processes. InAdvanced Informa- tion Systems Engineering, Matthias Jarke, John Mylopoulos, Christoph Quix, Co- lette Rolland, Yannis Manolopoulos, Haralamb...

    2014 doi

  16. [24]

    Nijat Mehdiyev, Maxim Majlatow, and Peter Fettke. 2024. Communicating un- certainty in machine learning explanations: a visualization analytics approach for predictive process monitoring. InWorld Conference on Explainable Artificial Intelligence. Springer, 420–438

    2024

  17. [25]

    Nijat Mehdiyev, Maxim Majlatow, and Peter Fettke. 2024. Counterfactual Ex- planations in the Big Picture: An Approach for Process Prediction-Driven Job- Shop Scheduling Optimization.Cognitive Computation16, 5 (2024), 2674–2700. doi:10.1007/s12559-024-10294-0

    2024 doi

  18. [26]

    Nijat Mehdiyev, Maxim Majlatow, and Peter Fettke. 2024. Quantifying and explaining machine learning uncertainty in predictive process monitoring: an operations research perspective.Annals of Operations Research(2024), 1–40. doi:10.1007/s10479-024-05943-4

    2024 doi

  19. [27]

    Nijat Mehdiyev, Maxim Majlatow, and Peter Fettke. 2025. Augmenting post-hoc explanations for predictive process monitoring with uncertainty quantification via conformalized Monte Carlo dropout.Data & Knowledge Engineering156 (2025), 102402. doi:10.1016/j.datak.2024.102402

    2025

  20. [28]

    Nijat Mehdiyev, Maxim Majlatow, and Peter Fettke. 2025. Integrating permutation feature importance with conformal prediction for robust Explainable Artificial Intelligence in predictive process monitoring.Engineering Applications of Artificial Intelligence149 (2025), 110363. d...

    2025

  21. [29]

    Martin Meitz, Henrik Leopold, and Jan Mendling. 2013. An Approach to Support Process Model Validation based on Text Generation.EMISA FORUM33, 2 (aug 2013), 7–20. doi:10.1007/BF03345896

    2013 doi

  22. [30]

    Tim Miller. 2023. Explainable AI is Dead, Long Live Explainable AI! Hypothesis- driven Decision Support using Evaluative AI. InProceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency(Chicago, IL, USA)(FAccT ’23). Association for Computing Machinery...

    2023

  23. [31]

    Sina Mohseni, Niloofar Zarei, and Eric D. Ragan. 2021. A Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems.ACM Trans. Interact. Intell. Syst.11, 3–4, Article 24 (Sept. 2021), 45 pages. doi:10.1145/3387166

    2021 doi

  24. [32]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Stefano Ermon, Christopher D Manning, and Chelsea Finn. 2023. Direct preference optimization: your language model is secretly a reward model. InProceedings of the 37th International Confer- ence on Neural Information Processing Sy...

    2023

  25. [33]

    Héctor Guadalupe Ramírez-Escamilla, María Concepción Martínez-Rodríguez, Alejandro Padilla-Rivera, Diego Domínguez-Solís, and Lorena Elizabeth Campos- Villegas. 2024. Advancing Toward Sustainability: A Systematic Review of Circular Economy Strategies in the Textile Industry.Re...

    2024

  26. [34]

    Williams Rizzi, Chiara Di Francescomarino, and Fabrizio Maria Maggi. 2020. Explainability in Predictive Process Monitoring: When Understanding Helps Improving. InBusiness Process Management Forum, Dirk Fahland, Chiara Ghidini, Jörg Becker, and Marlon Dumas (Eds.). Springer Int...

    2020 doi

  27. [35]

    Cynthia Rudin, Chaofan Chen, Zhi Chen, Haiyang Huang, Lesia Semenova, and Chudi Zhong. 2022. Interpretable machine learning: Fundamental principles and 10 grand challenges.Statistics Surveys16 (2022), 1 – 85. doi:10.1214/21-SS133

    2022 doi

  28. [36]

    Busra L Sebin, Nazim Ali Taskin, and Nijat Mehdiyev. 2024. Exploring the Intersection of Large Language Models (LLMs) and Explainable AI (XAI): A Systematic Literature Review (Research-in-Progress).CONF-IRM 2024 Proceedings. 25(2024)

    2024

  29. [37]

    Juan Sequeda, Dean Allemang, and Bryon Jacob. 2024. A Benchmark to Un- derstand the Role of Knowledge Graphs on Large Language Model’s Accuracy for Question Answering on Enterprise SQL Databases. InProceedings of the 7th Joint Workshop on Graph Data Management Experiences & Sy...

    2024

  30. [38]

    Dylan Slack, Satyapriya Krishna, Himabindu Lakkaraju, and Sameer Singh. 2023. Explaining machine learning models with interactive natural language conver- sations using TalkToModel.Nature Machine Intelligence5, 8 (2023), 873–883. doi:10.1038/s42256-023-00692-8 2025 ACM SIGMOD/...

    2023 doi

  31. [39]

    Irene Teinemaa, Marlon Dumas, Marcello La Rosa, and Fabrizio Maria Maggi

  32. [40]

    Wil Van Der Aalst. 2012. Process mining.Commun. ACM55, 8 (Aug. 2012), 76–83. doi:10.1145/2240236.2240257

    2012

  33. [41]

    Vladimir Vovk, Ivan Petej, Ilia Nouretdinov, Valery Manokhin, and Alexander Gammerman. 2020. Computationally efficient versions of conformal predictive distributions.Neurocomputing397 (2020), 292–308. doi:10.1016/j.neucom.2019. 10.110

    2020 doi

  34. [42]

    Wenjie Wu, Yongcheng Jing, Yingjie Wang, Wenbin Hu, and Dacheng Tao. 2025. Graph-augmented reasoning: Evolving step-by-step knowledge graph retrieval for llm reasoning.arXiv preprint arXiv:2503.01642(2025). Received 15.05.2025

    2025 arXiv

  35. [2019]

    ACM Trans

    Outcome-Oriented Predictive Process Monitoring: Review and Benchmark. ACM Trans. Knowl. Discov. Data13, 2, Article 17 (March 2019), 57 pages. doi:10. 1145/3301300

    2019

  36. [2024]

    InEnterprise, Business- Process and Information Systems Modeling, Han van der Aa, Dominik Bork, Rainer Schmidt, and Arnon Sturm (Eds.)

    Process Modeling with Large Language Models. InEnterprise, Business- Process and Information Systems Modeling, Han van der Aa, Dominik Bork, Rainer Schmidt, and Arnon Sturm (Eds.). Springer Nature Switzerland, Cham, 229–244. doi:10.1007/978-3-031-61007-3_18

    doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.