Pith. sign in

REVIEW 2 major objections 1 minor 52 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

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

T0 review · grok-4.3

PropLLM traces network faults hop-by-hop from end alerts using LLMs guided by causal attention on a knowledge graph to build an evidenced root-cause chain.

2026-06-28 18:56 UTC pith:63CYIUW7

load-bearing objection PropLLM tries hop-by-hop reconstruction with LLMs and a causal attention trick for network faults, showing small gains on two datasets, but the abstract gives almost no implementation or baseline detail. the 2 major comments →

arxiv 2606.00582 v1 pith:63CYIUW7 submitted 2026-05-30 cs.AI

PropLLM: Propagation-Aware Scene Reconstruction for Network Fault Diagnosis

Zongzong Wu , Ming Zhao , Fengxiao Tang , Nei Kato This is my paper
classification cs.AI
keywords network fault diagnosislarge language modelsknowledge graphcausal propagationscene reconstructionroot cause localizationattention mechanism
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper sets out to establish that network fault diagnosis suffers from inherent ambiguity when systems see only the final symptoms, because multiple root faults produce similar end-point alerts. Single-pass mapping methods, whether rule-based, machine-learning, or LLM-based, cannot resolve this ambiguity because they lack a mechanism to walk the propagation path backward. PropLLM attempts to solve the problem by starting at the alerts, stepping hop-by-hop along the topology, pulling verifiable evidence from a dual-layer knowledge graph at every step, and using a Temporal Causal Propagation Attention layer to keep the LLM's reasoning aligned with known causal directions. If the method works, diagnosis produces an explicit, evidence-linked chain to both the fault location and the fault type rather than a direct guess from the symptom set.

Core claim

PropLLM is the first to integrate the hop-by-hop scene reconstruction paradigm with the generative reasoning capabilities of LLMs. Starting from end-point alerts, PropLLM traces back hop-by-hop along the propagation path, retrieving verifiable factual evidence from a dual-layer knowledge graph at each hop, while the proposed Temporal Causal Propagation Attention mechanism encodes known topological causal priors directly into the attention computation to guide the model along the correct causal direction, ultimately localizing the root cause and determining the fault type through a fully evidenced causal chain.

What carries the argument

The Temporal Causal Propagation Attention (TCPA) mechanism, which injects known topological causal priors into the LLM attention computation to steer reasoning along the actual propagation direction at each hop.

Load-bearing premise

That a dual-layer knowledge graph supplies accurate, retrievable evidence for each hop along the true propagation path starting from the observed end-point alerts.

What would settle it

On the real-world Wi-Fi multimodal fault dataset, running the model without the hop-by-hop tracing step or without the TCPA mechanism yields no improvement in root-cause localization accuracy over the strongest single-pass baseline.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Fault-type diagnosis accuracy rises 3.9 percent and root-cause localization accuracy rises 4.7 percent over the strongest baseline on the Wi-Fi dataset.
  • Hallucination rate falls by 50.8 percent because each reasoning step is anchored to retrieved factual evidence.
  • The same gains appear on the TeleLogs 5G dataset, indicating the approach transfers across network types.
  • Single-pass methods remain structurally unable to resolve end-point ambiguity because they never reconstruct the intermediate causal steps.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same hop-by-hop evidence-retrieval pattern could be applied to any domain in which effects propagate through a known topology, such as supply-chain disruption tracing or epidemiological contact tracing.
  • If the dual-layer knowledge graph can be kept current, the method might support online diagnosis rather than post-hoc analysis.
  • The reduction in hallucinations through per-hop evidence checks may generalize to other LLM tasks that require long causal chains in technical domains.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper proposes PropLLM, which integrates hop-by-hop scene reconstruction with LLM generative reasoning for network fault diagnosis. Starting from end-point alerts, it traces propagation paths using a dual-layer knowledge graph for verifiable evidence at each hop and introduces a Temporal Causal Propagation Attention (TCPA) mechanism to encode topological causal priors into attention computations. This produces a fully evidenced causal chain for root cause localization and fault type determination. Experiments on a real-world Wi-Fi multimodal fault dataset report 3.9% higher fault type diagnosis accuracy, 4.7% higher root cause localization accuracy, and 50.8% lower hallucination rate versus the strongest baseline, with supplementary results on the TeleLogs 5G dataset.

Significance. If the hop-by-hop reconstruction and TCPA-guided reasoning hold up under detailed scrutiny, the work could meaningfully advance LLM applications in operations by replacing single-pass alert-to-diagnosis mappings with traceable causal chains, potentially lowering hallucination rates in safety-critical network settings. The use of real-world multimodal and 5G datasets is a positive factor for practical relevance.

major comments (2)
  1. [Abstract] Abstract: The reported accuracy gains (3.9%, 4.7%) and hallucination reduction (50.8%) are presented without baseline descriptions, dataset statistics, error bars, or statistical significance tests. This information is load-bearing for assessing whether the central claim of effectiveness is supported.
  2. [Introduction] The assertion that rule-based, ML-based, and LLM-based methods are 'structurally incapable' of resolving end-point ambiguity is central to the motivation but lacks a concrete comparison (e.g., why a suitably prompted LLM cannot perform multi-hop reasoning in one pass). This needs explicit treatment in the related-work or method sections.
minor comments (1)
  1. [Abstract] The dual-layer knowledge graph is referenced repeatedly but its construction, schema, and update mechanism receive no high-level description in the abstract or early sections, complicating reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and will revise the paper accordingly to strengthen the presentation of results and motivation.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The reported accuracy gains (3.9%, 4.7%) and hallucination reduction (50.8%) are presented without baseline descriptions, dataset statistics, error bars, or statistical significance tests. This information is load-bearing for assessing whether the central claim of effectiveness is supported.

    Authors: We agree that the abstract would benefit from additional context on the reported metrics. In the revised manuscript, we will update the abstract to name the strongest baseline (the best-performing prior method from our experiments), include brief dataset statistics (e.g., number of samples and fault types in the Wi-Fi and TeleLogs datasets), and reference that error bars and statistical significance tests appear in Section 5. Due to strict length limits on abstracts, we cannot embed full tables or p-values there, but the added phrasing will make the claims self-contained while directing readers to the detailed experimental analysis. revision: partial

  2. Referee: [Introduction] The assertion that rule-based, ML-based, and LLM-based methods are 'structurally incapable' of resolving end-point ambiguity is central to the motivation but lacks a concrete comparison (e.g., why a suitably prompted LLM cannot perform multi-hop reasoning in one pass). This needs explicit treatment in the related-work or method sections.

    Authors: We accept that the structural-incapability claim requires explicit justification. In the revised manuscript we will insert a new subsection in Related Work (or an expanded paragraph in Section 3) that contrasts single-pass LLM prompting with our approach. Specifically, we will explain that even chain-of-thought or multi-turn prompting on a single forward pass cannot guarantee retrieval of verifiable KG evidence at each hop nor enforce the TCPA causal priors; we will support this with a short qualitative example drawn from the Wi-Fi dataset showing how one-pass models produce ungrounded hops while PropLLM's iterative reconstruction avoids them. This addition will be placed before the method description. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper presents an architectural system (hop-by-hop reconstruction + LLM + dual-layer KG + TCPA attention) whose central claims are descriptive and empirical rather than derived from equations or fitted parameters. No load-bearing step reduces by construction to its own inputs, no self-citation chain is invoked to justify uniqueness, and no predictions are statistically forced from fitted quantities. The reported accuracy gains on external datasets are independent of any internal definitional loop.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Ledger extracted from abstract only; full methods section unavailable.

axioms (1)
  • domain assumption The dual-layer knowledge graph contains accurate, verifiable factual evidence for propagation paths at each hop.
    Method depends on retrieving such evidence to build the causal chain.
invented entities (1)
  • Temporal Causal Propagation Attention (TCPA) no independent evidence
    purpose: Encodes topological causal priors directly into attention computation to guide the model along the correct causal direction.
    New mechanism introduced to steer LLM reasoning.

pith-pipeline@v0.9.1-grok · 5785 in / 1128 out tokens · 26852 ms · 2026-06-28T18:56:11.748391+00:00 · methodology

0 comments
read the original abstract

Network faults propagate layer by layer along topology and protocol dependencies, yet operations systems typically observe only symptomatic alerts at the tail end of propagation chains, where distinct root-cause faults may produce highly similar end-point symptoms. Existing approaches, whether rule-based, machine learning (ML)-based, or large language model (LLM)-based, fundamentally map the alert set to a diagnosis in a single pass and are structurally incapable of resolving this end-point ambiguity. This paper proposes PropLLM, which is the first to integrate the hop-by-hop scene reconstruction paradigm with the generative reasoning capabilities of LLMs. Starting from end-point alerts, PropLLM traces back hop-by-hop along the propagation path, retrieving verifiable factual evidence from a dual-layer knowledge graph (KG) at each hop, while the proposed Temporal Causal Propagation Attention (TCPA) mechanism encodes known topological causal priors directly into the attention computation to guide the model along the correct causal direction, ultimately localizing the root cause and determining the fault type through a fully evidenced causal chain. On a real-world Wi-Fi multimodal fault dataset, PropLLM improves fault type diagnosis accuracy by 3.9\% and root cause localization accuracy by 4.7\% over the strongest baseline, while reducing the hallucination rate by 50.8\%. Supplementary experiments on the TeleLogs 5G dataset further demonstrate the effectiveness of the proposed method across different network scenarios.

Figures

Figures reproduced from arXiv: 2606.00582 by Fengxiao Tang, Ming Zhao, Nei Kato, Zongzong Wu.

Figure 1
Figure 1. Figure 1: (a) Fault cascading propagation produces ambiguous end-point alerts. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of the dual-layer knowledge graph with cross-layer [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overall architecture of the PropLLM framework. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Conditional association between root cause localization correctness [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Multi-dimensional performance radar chart of PropLLM and repre [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Attention weight matrices of standard self-attention (left) vs. TCPA [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Confusion matrix heatmaps of BiAn (left) and PropLLM (right). [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Misclassification counts on top-5 confusion pairs. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

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

52 extracted references · 5 canonical work pages · 2 internal anchors

  1. [1]

    A survey of fault localization techniques in computer networks,

    M. łgorzata Steinder and A. S. Sethi, “A survey of fault localization techniques in computer networks,”Science of computer programming, vol. 53, no. 2, pp. 165–194, 2004

    2004

  2. [2]

    A survey of aiops for failure management in the era of large language models,

    L. Zhang, T. Jia, M. Jia, Y . Wu, A. Liu, Y . Yang, Z. Wu, X. Hu, P. S. Yu, and Y . Li, “A survey of aiops for failure management in the era of large language models,”arXiv preprint arXiv:2406.11213, 2024

    work page arXiv 2024

  3. [3]

    Deep learning and its applications to machine health monitoring,

    R. Zhao, R. Yan, Z. Chen, K. Mao, P. Wang, and R. X. Gao, “Deep learning and its applications to machine health monitoring,”Mechanical systems and signal processing, vol. 115, pp. 213–237, 2019

    2019

  4. [4]

    Graph neural network-based anomaly detection in multivariate time series,

    A. Deng and B. Hooi, “Graph neural network-based anomaly detection in multivariate time series,” inProceedings of the AAAI conference on artificial intelligence, vol. 35, no. 5, 2021, pp. 4027–4035

    2021

  5. [5]

    Recommending root-cause and mitigation steps for cloud incidents using large language models,

    T. Ahmed, S. Ghosh, C. Bansal, T. Zimmermann, X. Zhang, and S. Rajmohan, “Recommending root-cause and mitigation steps for cloud incidents using large language models,” in2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE). IEEE, 2023, pp. 1737–1749

    2023

  6. [6]

    Interleav- ing retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions,

    H. Trivedi, N. Balasubramanian, T. Khot, and A. Sabharwal, “Interleav- ing retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions,” inProceedings of the 61st annual meeting of the association for computational linguistics (volume 1: long papers), 2023, pp. 10 014–10 037

    2023

  7. [7]

    Towards llm-based failure localization in production-scale networks,

    C. Wang, X. Zhang, R. Lu, X. Lin, X. Zeng, X. Zhang, Z. An, G. Wu, J. Gao, C. Tianet al., “Towards llm-based failure localization in production-scale networks,” inProceedings of the ACM SIGCOMM 2025 Conference, 2025, pp. 496–511

    2025

  8. [8]

    {NetAssistant}: Dialogue based network diagnosis in data center networks,

    H. Wang, A. Abhashkumar, C. Lin, T. Zhang, X. Gu, N. Ma, C. Wu, S. Liu, W. Zhou, Y . Donget al., “{NetAssistant}: Dialogue based network diagnosis in data center networks,” in21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), 2024, pp. 2011–2024

    2024

  9. [9]

    Automatic root cause analysis via large language models for cloud incidents,

    Y . Chen, H. Xie, M. Ma, Y . Kang, X. Gao, L. Shi, Y . Cao, X. Gao, H. Fan, M. Wenet al., “Automatic root cause analysis via large language models for cloud incidents,” inProceedings of the Nineteenth European Conference on Computer Systems, 2024, pp. 674–688

    2024

  10. [10]

    A survey on knowledge graphs: Representation, acquisition, and applications,

    S. Ji, S. Pan, E. Cambria, P. Marttinen, and P. S. Yu, “A survey on knowledge graphs: Representation, acquisition, and applications,”IEEE transactions on neural networks and learning systems, vol. 33, no. 2, pp. 494–514, 2021

    2021

  11. [11]

    A concurrent fault diagnosis method of transformer based on graph convolutional network and knowledge graph,

    L. Liu, B. Wang, F. Ma, Q. Zheng, L. Yao, C. Zhang, and M. A. Mohamed, “A concurrent fault diagnosis method of transformer based on graph convolutional network and knowledge graph,”Frontiers in Energy Research, vol. 10, p. 837553, 2022

    2022

  12. [12]

    A comprehensive survey on root cause analysis in (micro) services: methodologies, challenges, and trends,

    T. Wang and G. Qi, “A comprehensive survey on root cause analysis in (micro) services: Methodologies, challenges, and trends,”arXiv preprint arXiv:2408.00803, 2024

    work page arXiv 2024

  13. [13]

    Heterogeneous graph trans- former,

    Z. Hu, Y . Dong, K. Wang, and Y . Sun, “Heterogeneous graph trans- former,” inProceedings of the web conference 2020, 2020, pp. 2704– 2710

    2020

  14. [14]

    Retrieval- augmented generation for knowledge-intensive nlp tasks,

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W.-t. Yih, T. Rockt ¨aschelet al., “Retrieval- augmented generation for knowledge-intensive nlp tasks,”Advances in neural information processing systems, vol. 33, pp. 9459–9474, 2020

    2020

  15. [15]

    Self-rag: Learning to retrieve, generate, and critique through self-reflection,

    A. Asai, Z. Wu, Y . Wang, A. Sil, and H. Hajishirzi, “Self-rag: Learning to retrieve, generate, and critique through self-reflection,” inInterna- tional conference on learning representations, vol. 2024, 2024, pp. 9112–9141

    2024

  16. [16]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  17. [17]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, F. Xia, E. Chi, Q. V . Le, D. Zhouet al., “Chain-of-thought prompting elicits reasoning in large language models,”Advances in neural information processing systems, vol. 35, pp. 24 824–24 837, 2022

    2022

  18. [18]

    Root cause analysis of failures in microservices through causal discovery,

    A. Ikram, S. Chakraborty, S. Mitra, S. Saini, S. Bagchi, and M. Ko- caoglu, “Root cause analysis of failures in microservices through causal discovery,”Advances in Neural Information Processing Systems, vol. 35, pp. 31 158–31 170, 2022

    2022

  19. [19]

    Nezha: Interpretable fine-grained root causes analysis for microservices on multi-modal observability data,

    G. Yu, P. Chen, Y . Li, H. Chen, X. Li, and Z. Zheng, “Nezha: Interpretable fine-grained root causes analysis for microservices on multi-modal observability data,” inProceedings of the 31st ACM joint European software engineering conference and symposium on the foun- dations of software engineering, 2023, pp. 553–565

    2023

  20. [20]

    Chain-of-event: Interpretable root cause analysis for microser- vices through automatically learning weighted event causal graph,

    Z. Yao, C. Pei, W. Chen, H. Wang, L. Su, H. Jiang, Z. Xie, X. Nie, and D. Pei, “Chain-of-event: Interpretable root cause analysis for microser- vices through automatically learning weighted event causal graph,” in Companion Proceedings of the 32nd ACM International Conference on the Foundations of Software Engineering, 2024, pp. 50–61

    2024

  21. [21]

    {ExChain}: Exception dependency analysis for root cause diagnosis,

    A. Li, S. Lu, S. Nath, R. Padhye, and V . Sekar, “{ExChain}: Exception dependency analysis for root cause diagnosis,” in21st USENIX Sym- posium on Networked Systems Design and Implementation (NSDI 24), 2024, pp. 2047–2062

    2024

  22. [22]

    Illuminating the gray zone: Non- intrusive gray failure localization in server operating systems,

    S. Zhang, Y . Zhao, X. Xiong, Y . Sun, X. Nie, J. Zhang, F. Wang, X. Zheng, Y . Zhang, and D. Pei, “Illuminating the gray zone: Non- intrusive gray failure localization in server operating systems,” inCom- panion Proceedings of the 32nd ACM International Conference on the Foundations of Software Engineering, 2024, pp. 126–137

    2024

  23. [23]

    Mulan: Multi-modal causal structure learning and root cause analysis for microservice systems,

    L. Zheng, Z. Chen, J. He, and H. Chen, “Mulan: Multi-modal causal structure learning and root cause analysis for microservice systems,” in Proceedings of the ACM Web Conference 2024, 2024, pp. 4107–4116

    2024

  24. [24]

    Minder: Faulty machine detection for large-scale distributed model training,

    Y . Deng, X. Shi, Z. Jiang, X. Zhang, L. Zhang, Z. Zhang, B. Li, Z. Song, H. Zhu, G. Liuet al., “Minder: Faulty machine detection for large-scale distributed model training,” in22nd USENIX Symposium on Networked Systems Design and Implementation (NSDI 25), 2025, pp. 505–521

    2025

  25. [25]

    Xpert: Empowering incident management with query recommendations via large language models,

    Y . Jiang, C. Zhang, S. He, Z. Yang, M. Ma, S. Qin, Y . Kang, Y . Dang, S. Rajmohan, Q. Linet al., “Xpert: Empowering incident management with query recommendations via large language models,” inProceedings of the IEEE/ACM 46th International conference on software engineering, 2024, pp. 1–13

    2024

  26. [26]

    Lm-pace: Confidence estimation by large language models for effective root causing of cloud incidents,

    D. Zhang, X. Zhang, C. Bansal, P. Las-Casas, R. Fonseca, and S. Raj- mohan, “Lm-pace: Confidence estimation by large language models for effective root causing of cloud incidents,” inCompanion Proceedings of the 32nd ACM International Conference on the Foundations of Software Engineering, 2024, pp. 388–398

    2024

  27. [27]

    No more data silos: Unified microservice failure diagnosis with temporal knowledge graph,

    S. Zhang, Y . Zhao, S. Xia, S. Wei, Y . Sun, C. Zhao, S. Ma, J. Kuang, B. Zhu, L. Panet al., “No more data silos: Unified microservice failure diagnosis with temporal knowledge graph,”IEEE Transactions on Services Computing, vol. 17, no. 6, pp. 4013–4026, 2024

    2024

  28. [28]

    Kg4diagnosis: A hierarchical multi- agent llm framework with knowledge graph enhancement for medical diagnosis,

    K. Zuo, Y . Jiang, F. Mo, and P. Lio, “Kg4diagnosis: A hierarchical multi- agent llm framework with knowledge graph enhancement for medical diagnosis,” inAAAI Bridge Program on AI for Medicine and Healthcare. PMLR, 2025, pp. 195–204

    2025

  29. [29]

    Kgv: Integrating large language models with knowledge graphs for cyber threat intelligence credibility assessment,

    Z. Wu, M. Zhao, Y . Li, H. Li, S. Peng, N. Kato, and F. Tang, “Kgv: Integrating large language models with knowledge graphs for cyber threat intelligence credibility assessment,”IEEE Transactions on Mobile Computing, 2026

    2026

  30. [30]

    Reasoning on graphs: Faithful and interpretable large language model reasoning,

    L. Luo, Y .-F. Li, R. Haffari, and S. Pan, “Reasoning on graphs: Faithful and interpretable large language model reasoning,” inInternational Conference on Learning Representations, vol. 2024, 2024, pp. 14 400– 14 423

    2024

  31. [31]

    Kirag: Knowledge-driven iterative retriever for enhancing retrieval-augmented generation,

    J. Fang, Z. Meng, and C. Macdonald, “Kirag: Knowledge-driven iterative retriever for enhancing retrieval-augmented generation,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025, pp. 18 969–18 985

    2025

  32. [32]

    Causal- former: An interpretable transformer for temporal causal discovery,

    L. Kong, W. Li, H. Yang, Y . Zhang, J. Guan, and S. Zhou, “Causal- former: An interpretable transformer for temporal causal discovery,” IEEE Transactions on Knowledge and Data Engineering, vol. 37, no. 1, pp. 102–115, 2024

    2024

  33. [33]

    Teaching transformers causal rea- soning through axiomatic training,

    A. Vashishtha, A. Kumar, A. Pandey, A. G. Reddy, K. Ahuja, V . N. Balasubramanian, and A. Sharma, “Teaching transformers causal rea- soning through axiomatic training,” inProc. International Conference on Machine Learning (ICML), 2025

    2025

  34. [34]

    Cuts+: High- dimensional causal discovery from irregular time-series,

    Y . Cheng, L. Li, T. Xiao, Z. Li, J. Suo, K. He, and Q. Dai, “Cuts+: High- dimensional causal discovery from irregular time-series,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 10, 2024, pp. 11 525–11 533

    2024

  35. [35]

    Causal interpretation of self-attention in pre-trained transformers,

    R. Y . Rohekar, Y . Gurwicz, and S. Nisimov, “Causal interpretation of self-attention in pre-trained transformers,”Advances in Neural Informa- tion Processing Systems, vol. 36, pp. 31 450–31 465, 2023

    2023

  36. [36]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” inPro- ceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), 2019, pp. 4171–4186

    2019

  37. [37]

    Long short-term memory,

    J. Schmidhuber, S. Hochreiteret al., “Long short-term memory,”Neural Comput, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  38. [38]

    Graph attention networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y . Bengio et al., “Graph attention networks,” inInternational conference on learning representations, vol. 6, no. 2. Ithaca, 2018

    2018

  39. [39]

    Kgpt: Knowledge-grounded pre-training for data-to-text generation,

    W. Chen, Y . Su, X. Yan, and W. Y . Wang, “Kgpt: Knowledge-grounded pre-training for data-to-text generation,” inProceedings of the 2020 con- ference on empirical methods in natural language processing (EMNLP), 2020, pp. 8635–8648

    2020

  40. [40]

    Diffusion improves graph learning,

    J. Gasteiger, S. Weißenberger, and S. G ¨unnemann, “Diffusion improves graph learning,”Advances in neural information processing systems, vol. 32, 2019

    2019

  41. [41]

    On layer normalization in the transformer architecture,

    R. Xiong, Y . Yang, D. He, K. Zheng, S. Zheng, C. Xing, H. Zhang, Y . Lan, L. Wang, and T. Liu, “On layer normalization in the transformer architecture,” inInternational conference on machine learning. PMLR, 2020, pp. 10 524–10 533

    2020

  42. [42]

    Toward Realistic Wi-Fi Fault Diagnosis: A Multi-Modal Benchmark

    J. Zhang, H. Deng, X. Li, M. Zhao, F. Tang, and N. Kato, “Toward realistic wi-fi fault diagnosis: A multi-modal benchmark,” 2026. [Online]. Available: https://arxiv.org/abs/2605.22008

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  43. [43]

    Reasoning Language Models for Root Cause Analysis in 5G Wireless Networks.arXiv preprint arXiv:2507.21974, 2025

    M. Sana, N. Piovesan, A. De Domenico, Y . Kang, H. Zhang, M. Debbah, and F. Ayed, “Reasoning language models for root cause analysis in 5g wireless networks,”arXiv preprint arXiv:2507.21974, 2025

    work page arXiv 2025

  44. [44]

    Qwen3 Technical Report

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lvet al., “Qwen3 technical report,”arXiv preprint arXiv:2505.09388, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  45. [45]

    Xgboost: A scalable tree boosting system,

    T. Chen and C. Guestrin, “Xgboost: A scalable tree boosting system,” inProceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 2016, pp. 785–794

    2016

  46. [46]

    Famos: Fault diagnosis for microservice systems through effective multi-modal data fusion,

    C. Duan, Y . Yang, T. Jia, G. Liu, J. Liu, H. Zhang, Q. Zhou, Y . Li, and G. Huang, “Famos: Fault diagnosis for microservice systems through effective multi-modal data fusion,” in2025 IEEE/ACM 47th Interna- tional Conference on Software Engineering (ICSE). IEEE, 2025, pp. 2613–2624

    2025

  47. [47]

    Netllm: Adapting large language models for networking,

    D. Wu, X. Wang, Y . Qiao, Z. Wang, J. Jiang, S. Cui, and F. Wang, “Netllm: Adapting large language models for networking,” inProceed- ings of the ACM SIGCOMM 2024 Conference, 2024, pp. 661–678

    2024

  48. [48]

    Intent-driven network management with multi-agent llms: The confucius framework,

    Z. Wang, S. Lin, G. Yan, S. Ghorbani, M. Yu, J. Zhou, N. Hu, L. Baruah, S. Peters, S. Kamathet al., “Intent-driven network management with multi-agent llms: The confucius framework,” inProceedings of the ACM SIGCOMM 2025 Conference, 2025, pp. 347–362

    2025

  49. [49]

    Incremental causal graph learning for online root cause analysis,

    D. Wang, Z. Chen, Y . Fu, Y . Liu, and H. Chen, “Incremental causal graph learning for online root cause analysis,” inProceedings of the 29th ACM SIGKDD conference on knowledge discovery and data mining, 2023, pp. 2269–2278

    2023

  50. [50]

    Root cause analysis in microservice using neural granger causal discovery,

    C.-M. Lin, C. Chang, W.-Y . Wang, K.-D. Wang, and W.-C. Peng, “Root cause analysis in microservice using neural granger causal discovery,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 1, 2024, pp. 206–213

    2024

  51. [51]

    Baro: Robust root cause analysis for microservices via multivariate bayesian online change point detection,

    L. Pham, H. Ha, and H. Zhang, “Baro: Robust root cause analysis for microservices via multivariate bayesian online change point detection,” Proceedings of the ACM on Software Engineering, vol. 1, no. FSE, pp. 2214–2237, 2024

    2024

  52. [52]

    Root cause analysis of anoma- lies in multivariate time series through granger causal discovery,

    X. Han, S. Absar, L. Zhang, and S. Yuan, “Root cause analysis of anoma- lies in multivariate time series through granger causal discovery,” inThe Thirteenth International Conference on Learning Representations, 2025

    2025