pith. sign in

arxiv: 2604.11040 · v1 · submitted 2026-04-13 · 💻 cs.AI

Intelligent Approval of Access Control Flow in Office Automation Systems via Relational Modeling

Dugang Liu , Zulong Chen , Chuanfei Xu , Jiaxuan He , Yunlu Ma , Jia Xu This is my paper

Pith reviewed 2026-05-10 15:59 UTC · model grok-4.3

classification 💻 cs.AI
keywords access controloffice automationrelational modelingintelligent approvalbinary relationsternary relationsACFA
0
0 comments X

The pith

Fusing binary applicant-approver relations with ternary applicant-resource-approver relations automates access control approvals in office systems.

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

The paper introduces the RMIA framework to replace step-by-step human approvals in office automation access control. It builds a binary module that captures coarse couplings between who requests and who approves, plus a ternary module that adds resource details for finer relations among requester, item, and approver. These two information streams are then combined into a single automated decision. The approach is tested on internal product data and live A/B trials to show it can cut manual effort while maintaining approval quality.

Core claim

RMIA consists of a binary relation modeling module that characterizes the coupling between applicants and approvers from a coarse-grained perspective and a ternary relation modeling module that uses resource information to capture complex relations among applicants, resources, and approvers from a fine-grained perspective; fusing the outputs of these modules produces the final automated approval decision for access control flow.

What carries the argument

RMIA framework that fuses a binary relation module (applicant-approver coupling) with a ternary relation module (applicant-resource-approver triples) to produce the approval decision.

Load-bearing premise

The fused binary and ternary models reliably reproduce the approval logic that human managers actually use across real requests.

What would settle it

A large-scale side-by-side comparison in which RMIA and human approvers independently score the same set of access requests; if the automated decisions diverge from human ones on a substantial fraction of cases or show consistent bias, the claim fails.

Figures

Figures reproduced from arXiv: 2604.11040 by Chuanfei Xu, Dugang Liu, Jia Xu, Jiaxuan He, Yunlu Ma, Zulong Chen.

Figure 1
Figure 1. Figure 1: An example of an access control flow approval [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The architecture of our Relational Modeling-driven Intelligent Approval (RMIA) framework, which primarily includes [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example of a ternary relation extractor. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: AUC results of ablation study on the Business-2 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Overview of the online ACFA intelligent approval [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Overview of the number of permission requests [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
read the original abstract

Office automation (OA) systems play a crucial role in enterprise operations and management, with access control flow approval (ACFA) being a key component that manages the accessibility of various resources. However, traditional ACFA requires approval from the person in charge at each step, which consumes a significant amount of manpower and time. Its intelligence is a crucial issue that needs to be addressed urgently by all companies. In this paper, we propose a novel relational modeling-driven intelligent approval (RMIA) framework to automate ACFA. Specifically, our RMIA consists of two core modules: (1) The binary relation modeling module aims to characterize the coupling relation between applicants and approvers and provide reliable basic information for ACFA decision-making from a coarse-grained perspective. (2) The ternary relation modeling module utilizes specific resource information as its core, characterizing the complex relations between applicants, resources, and approvers, and thus provides fine-grained gain information for informed decision-making. Then, our RMIA effectively fuses these two kinds of information to form the final decision. Finally, extensive experiments are conducted on two product datasets and an online A/B test to verify the effectiveness of RMIA.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 0 minor

Summary. The paper proposes the RMIA framework to automate access control flow approval (ACFA) in office automation systems. RMIA comprises a binary relation modeling module that captures coarse-grained applicant-approver couplings and a ternary relation modeling module that incorporates resource information for fine-grained applicant-resource-approver relations; these are fused to produce final approval decisions. Effectiveness is asserted via experiments on two product datasets and an online A/B test.

Significance. If the fusion of the binary and ternary modules reliably reproduces human approver logic across real requests without systematic bias or omitted policy factors, the work could deliver substantial practical value by reducing manual effort in enterprise access management. The use of product datasets and an A/B test provides a direct path to real-world validation, which is a positive aspect for applied systems research.

major comments (3)
  1. [Abstract] Abstract: the description states that RMIA 'effectively fuses these two kinds of information to form the final decision' yet supplies no equations, graph-construction procedure, embedding method, or fusion operator (e.g., attention, concatenation+MLP, or rule injection). This omission is load-bearing for the central claim that the model captures human approver logic.
  2. [Abstract] Abstract (and implied Methods): no details are given on the supervision signal, loss function, handling of conflicting approvals, or feature sets used in the relational modules. Without these, it cannot be verified whether key enterprise policy factors are omitted, directly engaging the stress-test concern that the fusion may fail to match human decisions.
  3. [Abstract] Abstract: the claim of verification via 'extensive experiments on two product datasets and an online A/B test' is unsupported by any reported metrics, baselines, error analysis, or statistical significance tests. This prevents evaluation of whether the binary/ternary fusion actually improves over simpler baselines.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment point by point below and agree that the abstract can be strengthened to better convey the technical details of RMIA.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the description states that RMIA 'effectively fuses these two kinds of information to form the final decision' yet supplies no equations, graph-construction procedure, embedding method, or fusion operator (e.g., attention, concatenation+MLP, or rule injection). This omission is load-bearing for the central claim that the model captures human approver logic.

    Authors: We agree that the abstract, as a concise summary, omits the specific technical implementation details such as equations, graph construction, embeddings, and the fusion operator. These aspects are described in the Methods section of the full manuscript. To address the concern, we will revise the abstract to include a brief high-level description of the relational modeling procedures and the fusion process used to combine the binary and ternary information. revision: yes

  2. Referee: [Abstract] Abstract (and implied Methods): no details are given on the supervision signal, loss function, handling of conflicting approvals, or feature sets used in the relational modules. Without these, it cannot be verified whether key enterprise policy factors are omitted, directly engaging the stress-test concern that the fusion may fail to match human decisions.

    Authors: We acknowledge that the abstract does not specify the supervision signal, loss function, handling of conflicting approvals, or feature sets. The full manuscript details these in the Methods section, including the use of historical approval records for supervision and the feature representations for applicants, resources, and approvers. We will revise the abstract to summarize the training supervision and loss to make these aspects clearer. revision: yes

  3. Referee: [Abstract] Abstract: the claim of verification via 'extensive experiments on two product datasets and an online A/B test' is unsupported by any reported metrics, baselines, error analysis, or statistical significance tests. This prevents evaluation of whether the binary/ternary fusion actually improves over simpler baselines.

    Authors: We agree that the abstract states the experimental validation without reporting specific metrics, baselines, or significance tests. The full manuscript includes these details in the Experiments section, with comparisons to baselines and results from the A/B test. We will revise the abstract to include key performance metrics and a note on the observed improvements to better support the effectiveness claim. revision: yes

Circularity Check

0 steps flagged

No circularity: RMIA is a modeling proposal verified by external experiments, with no self-referential derivations or fitted predictions.

full rationale

The paper introduces RMIA as a framework with binary and ternary relational modules whose outputs are fused for ACFA decisions, then validated on two product datasets plus an online A/B test. No equations, parameter-fitting steps, uniqueness theorems, or self-citations appear in the provided text that would make any claimed result equivalent to its inputs by construction. The central claim is an empirical modeling choice whose correctness is asserted via held-out data and live testing rather than by definitional reduction or self-referential citation chains. This is the expected non-circular outcome for an applied ML systems paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review prevents enumeration of specific free parameters or axioms; the framework implicitly assumes relational modeling suffices to encode approval logic and that product datasets reflect real decision distributions.

pith-pipeline@v0.9.0 · 5515 in / 955 out tokens · 36791 ms · 2026-05-10T15:59:24.885874+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

    Renuka Bhuyar and Saniya Ansari. 2016. Design and implementation of smart office automation system.International Journal of Computer Applications151, 3 (2016), 37–42

    work page 2016

  2. [2]

    Leo Breiman. 2001. Random forests.Machine Learning45, 1 (2001), 5–32

    work page 2001

  3. [3]

    Ning Chen, Bernardete Ribeiro, and An Chen. 2016. Financial credit risk assess- ment: A recent review.Artificial Intelligence Review45, 1 (2016), 1–23. Intelligent Approval of Access Control Flow in Office Automation Systems via Relational Modeling KDD ’26, August 9–13, 2026, Jeju, Korea

    work page 2016

  4. [4]

    Tianqi Chen and Carlos Guestrin. 2016. XGBoost: A scalable tree boosting system. InProceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 785–794

    work page 2016

  5. [5]

    Crispin Coombs, Donald Hislop, Stanimira K Taneva, and Sarah Barnard. 2020. The strategic impacts of Intelligent Automation for knowledge and service work: An interdisciplinary review.The Journal of Strategic Information Systems29, 4 (2020), 101600

    work page 2020

  6. [6]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the Association for Computational Linguistics. 4171–4186

    work page 2019

  7. [7]

    Muhammad K Farooq, Malik Jahan Khan, Shafay Shamail, and Mian M Awais

    work page

  8. [8]

    InProceedings of the 3rd International Conference on Theory and Practice of Electronic Governance

    Intelligent project approval cycle for local government: Case-based reason- ing approach. InProceedings of the 3rd International Conference on Theory and Practice of Electronic Governance. 68–73

    work page

  9. [9]

    Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: A factorization-machine based neural network for CTR prediction. In Proceedings of the 26th International Joint Conference on Artificial Intelligence. 1725–1731

    work page 2017

  10. [10]

    Peter Hofmann, Caroline Samp, and Nils Urbach. 2020. Robotic process automa- tion.Electronic Markets30, 1 (2020), 99–106

    work page 2020

  11. [11]

    Marina Johnson, Abdullah Albizri, and Antoine Harfouche. 2023. Responsible artificial intelligence in healthcare: Predicting and preventing insurance claim denials for economic and social wellbeing.Information Systems Frontiers25, 6 (2023), 2179–2195

    work page 2023

  12. [12]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. LightGBM: A highly efficient gradient boosting decision tree. InProceedings of the 31st International Conference on Neural Information Processing Systems. 3149–3157

    work page 2017

  13. [13]

    Damian Kedziora and Sami Hyrynsalmi. 2023. Turning robotic process automa- tion into intelligent automation with machine learning. InProceedings of the 11th International Conference on Communities and Technologies. 1–5

    work page 2023

  14. [14]

    Arun Kumar, Neeran Karnik, and Girish Chafle. 2002. Context sensitivity in role-based access control.ACM SIGOPS Operating Systems Review36, 3 (2002), 53–66

    work page 2002

  15. [15]

    Kyungsik Lee, Hana Yoo, Sumin Shin, Wooyoung Kim, Yeonung Baek, Hyunjin Kang, Jaehyun Kim, and Kee-Eung Kim. 2024. A submodular optimization ap- proach to accountable loan approval. InProceedings of the 38th AAAI Conference on Artificial Intelligence, Vol. 38. 22761–22769

    work page 2024

  16. [16]

    Yi Long. 2024. On the Sphere of Applying AI in Government Approval Service: Analysis According to the Data of Submitting in Seconds and Getting Result in Seconds Services in Shenzhen City. InProceedings of the 25th Annual International Conference on Digital Government Research. 932–940

    work page 2024

  17. [17]

    Kelong Mao, Jieming Zhu, Liangcai Su, Guohao Cai, Yuru Li, and Zhenhua Dong

    work page

  18. [18]

    In Proceedings of the 37th AAAI conference on Artificial Intelligence

    FinalMLP: An enhanced two-stream MLP model for CTR prediction. In Proceedings of the 37th AAAI conference on Artificial Intelligence. 4552–4560

    work page

  19. [19]

    Vincenzo Moscato, Antonio Picariello, and Giancarlo Sperlí. 2021. A benchmark of machine learning approaches for credit score prediction.Expert Systems with Applications165 (2021), 113986

    work page 2021

  20. [20]

    Margrethe H Olson and Henry C Lucas Jr. 1982. The impact of office automation on the organization: Some implications for research and practice.Commun. ACM 25, 11 (1982), 838–847

    work page 1982

  21. [21]

    Rouna Meisda Paoki and Jimmy Moedjahedy. 2024. Artificial intelligence and automation in office administrative procedures: A systematic literature review. YUME: Journal of Management7, 2 (2024), 321–331

    work page 2024

  22. [22]

    Priyadarshini R Pennathur, Valerie Boksa, Arunkumar Pennathur, Andrew Kusiak, and Beth A Livingston. 2024. The future of office and administrative support occupations in the era of artificial intelligence: A state of the art review and future research directions.International Journal of Industrial Ergonomics104 (2024), 103665

    work page 2024

  23. [23]

    Erasmo Purificato, Flavio Lorenzo, Francesca Fallucchi, and Ernesto William De Luca. 2023. The use of responsible artificial intelligence techniques in the context of loan approval processes.International Journal of Human–Computer Interaction39, 7 (2023), 1543–1562

    work page 2023

  24. [24]

    Lior Rokach. 2010. Ensemble-based classifiers.Artificial Intelligence Review33, 1 (2010), 1–39

    work page 2010

  25. [25]

    Segev Shlomov, Sami Marreed, and Avi Yaeli. 2024. Towards a resilient intelligent automation system. InProceedings of the 33rd International Joint Conference on Artificial Intelligence. 8797–8800

    work page 2024

  26. [26]

    Hao Tang, Cheng Wang, Jianguo Zheng, and Changjun Jiang. 2024. Enabling graph neural networks for semi-supervised risk prediction in online credit loan services.ACM Transactions on Intelligent Systems and Technology15, 1 (2024), 1–24

    work page 2024

  27. [27]

    Zhen Tian, Ting Bai, Wayne Xin Zhao, Ji-Rong Wen, and Zhao Cao. 2023. Eulernet: Adaptive feature interaction learning via Euler’s formula for ctr prediction. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1376–1385

    work page 2023

  28. [28]

    Juliana Tolles and William J Meurer. 2016. Logistic regression: Relating patient characteristics to outcomes.The Journal of the American Medical Association316, 5 (2016), 533–534

    work page 2016

  29. [29]

    Wil Van Der Aalst. 2012. Process mining: Overview and opportunities.ACM Transactions on Management Information Systems3, 2 (2012), 1–17

    work page 2012

  30. [30]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. InProceedings of the 31st International Conference on Neural Information Processing Systems. 6000–6010

    work page 2017

  31. [31]

    Ruoxi Wang, Rakesh Shivanna, Derek Cheng, Sagar Jain, Dong Lin, Lichan Hong, and Ed Chi. 2021. DCN v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems. InProceedings of the Web Conference 2021. 1785–1797

    work page 2021

  32. [32]

    Penyu Zhao, Di Zhang, and Sushan Lei. 2024. Design and implementation of office automation management system based on spring boot. InProceedings of the 2024 Guangdong-Hong Kong-Macao Greater Bay Area International Conference on Education Digitalization and Computer Science. 228–233

    work page 2024

  33. [33]

    Jieming Zhu, Qinglin Jia, Guohao Cai, Quanyu Dai, Jingjie Li, Zhenhua Dong, Ruiming Tang, and Rui Zhang. 2023. FINAL: Factorized interaction layer for ctr prediction. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2006–2010

    work page 2023