pith. sign in

arxiv: 2606.09892 · v1 · pith:SD43VKK4new · submitted 2026-06-03 · 💻 cs.LG · stat.ME

LMT: A Bayesian Framework for Causal Discovery from Textual Alarm Records in Manufacturing Systems

Xiaofeng Xiao , Jianhong Chen , Qiuzhuang Sun , Naichen Shi , Xubo Yue This is my paper

Pith reviewed 2026-06-28 07:00 UTC · model grok-4.3

classification 💻 cs.LG stat.ME
keywords causal discoveryBayesian methodslarge language modelsalarm logsmanufacturing systemstemporal point processesevent data analysis
0
0 comments X

The pith

LMT builds a prior over causal graphs from LLM analysis of alarm text and refines it with timestamp likelihoods to recover which events trigger others.

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

The paper presents LMT as a Bayesian method that extracts semantic signals about possible causal links from the textual descriptions in alarm logs and turns them into a prior distribution over graphs. It then updates that prior using the observed timing of events modeled as a Poisson process. The goal is to produce causal graphs that reflect both the meaning in the text and the statistical evidence from when events actually occur. A reader would care because manufacturing systems generate large volumes of alarm records where knowing true causes helps with diagnosis and prevention, yet text alone risks mistaking frequent sequences for causation and pure timing data needs large samples to be reliable. Simulations indicate the combined approach improves recovery especially when the number of observed events is small.

Core claim

LMT first uses LLMs to extract semantic causal signals from event descriptions and constructs a prior distribution over causal graphs among event types or event clusters. It then incorporates temporal evidence through a Poisson-process-based likelihood, allowing the LLM-informed prior to be refined by timestamp-based statistical evidence. By integrating the textual and temporal information, LMT produces a causal graph that is both interpretable and data-supported.

What carries the argument

The Bayesian update step that takes an LLM-derived prior over causal graphs and multiplies it by a Poisson-process likelihood on observed timestamps.

If this is right

  • In small-sample alarm-event scenarios the recovered graphs are more accurate than those from text-only or time-only methods.
  • The output graphs remain directly interpretable because each edge can be traced back to the original event descriptions used in the prior.
  • The framework remains effective across varied data-generating processes in the simulations.
  • The LLM prior can be refined by the timing likelihood without the semantic patterns forcing incorrect causal directions.

Where Pith is reading between the lines

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

  • The same structure could be applied to other logged event streams such as IT system alerts or hospital incident records where both text and timestamps are available.
  • If the LLM prior reliably encodes causal direction rather than mere association, the method could lower the sample size needed for reliable causal discovery in many engineering settings.
  • A direct test would be to run LMT on real factory data that also contains documented interventions and verify whether the inferred edges align with the known interventions.

Load-bearing premise

LLM analysis of alarm text supplies a prior over causal graphs that is meaningfully better than a non-informative prior and can be corrected by timing data without the semantics introducing systematic errors that equate correlation with causation.

What would settle it

Generate synthetic alarm sequences from a known true causal graph, apply LMT and baseline methods that use only text or only timestamps, and check whether LMT recovers the true edges at higher precision and recall, especially when the number of events is small.

Figures

Figures reproduced from arXiv: 2606.09892 by Jianhong Chen, Naichen Shi, Qiuzhuang Sun, Xiaofeng Xiao, Xubo Yue.

Figure 1
Figure 1. Figure 1: Comparison between the ground-truth and inferred cluster-level directed adjacency matrices under [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison between the ground-truth and inferred cluster-level directed adjacency matrices under [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example images of the handler. The handler contains multiple mechanical and sensing modules [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Learned module-cluster DAG. Each node represents a functional module of the handler, and each [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
read the original abstract

Textual event records, such as alarm logs, have become an increasingly common data source in engineering and manufacturing systems. Beyond identifying correlations or recurring patterns, engineers are often interested in understanding which types of events causally trigger or influence other events during system operation. Textual event descriptions may contain semantic clues about such causal relationships, and recent large language models (LLMs) provide a promising tool for extracting these signals. However, relying solely on LLM-encoded textual information is insufficient for accurate causal discovery, since semantic patterns do not directly reveal causal mechanisms and may confuse causation with correlation or frequent sequential patterns. To address these challenges, we propose \textbf{LMT}, a Bayesian causal discovery framework for engineering event data that jointly leverages textual descriptions and timestamps. Specifically, LMT first uses LLMs to extract semantic causal signals from event descriptions and constructs a prior distribution over causal graphs among event types or event clusters. It then incorporates temporal evidence through a Poisson-process-based likelihood, allowing the LLM-informed prior to be refined by timestamp-based statistical evidence. By integrating the textual and temporal information, LMT produces a causal graph that is both interpretable and data-supported. Simulation studies show that the proposed framework is effective across different settings and is especially advantageous in small-sample alarm-event scenarios.

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 / 2 minor

Summary. The paper presents LMT, a Bayesian causal discovery framework for textual alarm records in manufacturing systems. It uses LLMs to extract semantic causal signals from event descriptions and construct a prior distribution over causal graphs among event types or clusters. This prior is then updated via a Poisson-process-based likelihood that incorporates timestamp evidence. The integrated model is claimed to yield interpretable, data-supported causal graphs, with simulation studies demonstrating effectiveness across settings and particular advantages in small-sample alarm-event scenarios.

Significance. If the central claims hold, the work could meaningfully advance causal discovery methods for engineering domains that rely on textual event logs. Combining LLM-derived semantic priors with temporal statistical evidence addresses the practical challenge of limited data in manufacturing alarm systems, where purely data-driven or purely semantic approaches each have known limitations.

major comments (3)
  1. [Abstract] Abstract: The abstract states that simulations demonstrate effectiveness, especially in small samples, but supplies no equations, model details, data exclusion rules, or performance metrics. Without these, it is impossible to verify whether the central claim that the integrated graph is both interpretable and data-supported is actually supported by the experiments.
  2. [Abstract] Abstract / prior construction: The LLM-encoded textual information is asserted to supply a prior that is meaningfully better than non-informative and can be refined by the Poisson likelihood without systematic bias from semantic patterns that confuse correlation with causation. No independent calibration of the prior against ground-truth causal edges is described, which is load-bearing for the small-sample advantage claim because the prior dominates when event counts are low.
  3. [Method] Likelihood and update: The Poisson-process likelihood is presented as an independent source of temporal evidence that corrects the LLM prior, yet no evidence is given that the prior itself is not derived from patterns (e.g., co-occurrence or temporal order) that overlap with the likelihood, raising the risk that the Bayesian update reinforces rather than removes bias.
minor comments (2)
  1. [Notation] Notation for the causal graph prior and the Poisson intensity parameters should be defined explicitly at first use to improve readability.
  2. [Abstract] The abstract mentions "event types or event clusters" but does not clarify how clustering is performed or whether it affects the causal graph construction.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We respond to each major comment below, indicating planned revisions where the manuscript will be updated.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The abstract states that simulations demonstrate effectiveness, especially in small samples, but supplies no equations, model details, data exclusion rules, or performance metrics. Without these, it is impossible to verify whether the central claim that the integrated graph is both interpretable and data-supported is actually supported by the experiments.

    Authors: We agree that the abstract is concise and omits specific details. The model equations and Poisson likelihood are derived in Section 3, while simulation metrics (precision, recall, F1 across sample sizes), data generation, and exclusion criteria appear in Section 4. To address the concern, we will revise the abstract to include a brief summary of the key metrics showing small-sample advantages. revision: yes

  2. Referee: [Abstract] Abstract / prior construction: The LLM-encoded textual information is asserted to supply a prior that is meaningfully better than non-informative and can be refined by the Poisson likelihood without systematic bias from semantic patterns that confuse correlation with causation. No independent calibration of the prior against ground-truth causal edges is described, which is load-bearing for the small-sample advantage claim because the prior dominates when event counts are low.

    Authors: The prior is formed exclusively from LLM prompts on event descriptions to extract semantic causal signals. We acknowledge that no standalone calibration experiment against ground-truth edges is reported. The small-sample gains are shown via end-to-end simulations comparing LMT to non-informative priors. We will add a clarifying paragraph in the discussion on prior construction and potential semantic biases. revision: partial

  3. Referee: [Method] Likelihood and update: The Poisson-process likelihood is presented as an independent source of temporal evidence that corrects the LLM prior, yet no evidence is given that the prior itself is not derived from patterns (e.g., co-occurrence or temporal order) that overlap with the likelihood, raising the risk that the Bayesian update reinforces rather than removes bias.

    Authors: The LLM component receives only textual descriptions and is prompted solely for semantic causal inferences; no timestamps or co-occurrence counts are provided to it. The Poisson likelihood uses timestamp sequences exclusively. This separation by input modality ensures independence. We will revise the method section to state this separation explicitly. revision: yes

Circularity Check

0 steps flagged

No circularity: framework integrates independent textual prior and temporal likelihood without self-referential reduction.

full rationale

The paper describes LMT as constructing an LLM-based prior from event descriptions and refining it via a separate Poisson-process likelihood on timestamps. No equations, fitted parameters renamed as predictions, or self-citations are exhibited that would make the final graph equivalent to its inputs by construction. The two information sources are presented as distinct, with the Bayesian update operating on them without load-bearing self-definition or ansatz smuggling. This is the standard case of a self-contained proposal whose validity rests on external validation rather than internal reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated. The framework implicitly relies on the untested premise that LLM semantic signals form a valid causal prior and that Poisson processes adequately model alarm timing dependencies.

pith-pipeline@v0.9.1-grok · 5769 in / 1220 out tokens · 23663 ms · 2026-06-28T07:00:52.500968+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

104 extracted references · 2 linked inside Pith

  1. [2]

    Advances in Neural Information Processing Systems , volume=

    Causal discovery from event sequences by local cause-effect attribution , author=. Advances in Neural Information Processing Systems , volume=

  2. [3]

    Transactions on Machine Learning Research , year=

    Causal reasoning and large language models: Opening a new frontier for causality , author=. Transactions on Machine Learning Research , year=

  3. [4]

    IEEE Transactions on Reliability , volume=

    Robust statistical modeling of heterogeneity for repairable systems using multivariate Gaussian convolution processes , author=. IEEE Transactions on Reliability , volume=. 2023 , publisher=

    2023

  4. [5]

    Proceedings of the 17th international conference on Software engineering , pages=

    Automating process discovery through event-data analysis , author=. Proceedings of the 17th international conference on Software engineering , pages=

  5. [6]

    IBM Journal of Research and Development , volume=

    Predictive maintenance based on event-log analysis: A case study , author=. IBM Journal of Research and Development , volume=. 2017 , publisher=

    2017

  6. [7]

    Engineering Structures , volume=

    Towards fully automated post-event data collection and analysis: Pre-event and post-event information fusion , author=. Engineering Structures , volume=. 2020 , publisher=

    2020

  7. [8]

    Computers & Industrial Engineering , volume=

    Data-driven root cause analysis via causal discovery using time-to-event data , author=. Computers & Industrial Engineering , volume=. 2024 , publisher=

    2024

  8. [9]

    Applied Soft Computing , volume=

    Systems engineering issues for industry applications of large language model , author=. Applied Soft Computing , volume=. 2024 , publisher=

    2024

  9. [10]

    INCOSE International Symposium , volume=

    Impact Analysis of using Natural Language Processing and Large Language Model on Automated Correction of Systems Engineering Requirements , author=. INCOSE International Symposium , volume=. 2024 , organization=

    2024

  10. [11]

    Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Adaptive event detection with time-varying poisson processes , author=. Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  11. [12]

    2010 , publisher=

    Modeling events in time using cascades of Poisson processes , author=. 2010 , publisher=

    2010

  12. [14]

    IEEE Transactions on Knowledge and Data Engineering , year=

    Llm-driven causal discovery via harmonized prior , author=. IEEE Transactions on Knowledge and Data Engineering , year=

  13. [16]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Uncovering bias paths with llm-guided causal discovery: An active learning and dynamic scoring approach , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  14. [17]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Mitigating hallucinations in large language models via causal reasoning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  15. [18]

    Computational Analysis of Storylines: Making Sense of Events , volume=

    Event causality , author=. Computational Analysis of Storylines: Making Sense of Events , volume=. 2021 , publisher=

    2021

  16. [19]

    ACM Computing Surveys , volume=

    Causal discovery from temporal data: An overview and new perspectives , author=. ACM Computing Surveys , volume=. 2024 , publisher=

    2024

  17. [20]

    IEEE transactions on visualization and computer graphics , volume=

    Visual causality analysis of event sequence data , author=. IEEE transactions on visualization and computer graphics , volume=. 2020 , publisher=

    2020

  18. [21]

    Proceedings of the 2018 ACM SIGSAC conference on computer and communications security , pages=

    Nodemerge: Template based efficient data reduction for big-data causality analysis , author=. Proceedings of the 2018 ACM SIGSAC conference on computer and communications security , pages=

    2018

  19. [22]

    LearnDA: Learnable knowledge-guided data augmentation for event causality identification , author=. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) , pages=

  20. [23]

    International conference on knowledge engineering and knowledge management , pages=

    Beyond causality: Representing event relations in knowledge graphs , author=. International conference on knowledge engineering and knowledge management , pages=. 2022 , organization=

    2022

  21. [24]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Tnpar: Topological neural poisson auto-regressive model for learning granger causal structure from event sequences , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  22. [25]

    IEEE Transactions on Information Theory , year=

    A Poisson decomposition for information and the information-event diagram , author=. IEEE Transactions on Information Theory , year=

  23. [26]

    IEEE Internet of Things Journal , year=

    ANHP: Adaptive Neural Hawkes Processes for Causal Structure Learning on Event Sequences , author=. IEEE Internet of Things Journal , year=

  24. [27]

    Advances in Neural Information Processing Systems , volume=

    Dibs: Differentiable bayesian structure learning , author=. Advances in Neural Information Processing Systems , volume=

  25. [28]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Distributed representation of words in cause and effect spaces , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  26. [29]

    International conference on machine learning , pages=

    Learning granger causality for hawkes processes , author=. International conference on machine learning , pages=. 2016 , organization=

    2016

  27. [30]

    IEEE Transactions on Neural Networks and Learning Systems , volume=

    THPs: Topological Hawkes processes for learning causal structure on event sequences , author=. IEEE Transactions on Neural Networks and Learning Systems , volume=. 2022 , publisher=

    2022

  28. [31]

    Advances in neural information processing systems , volume=

    Learning hawkes processes from a handful of events , author=. Advances in neural information processing systems , volume=

  29. [32]

    Feature engineering for machine learning and data analytics , pages=

    Feature engineering for text data , author=. Feature engineering for machine learning and data analytics , pages=. 2018 , publisher=

    2018

  30. [33]

    Computers & Industrial Engineering , volume=

    Cost-based text understanding to improve maintenance knowledge intelligence in manufacturing enterprises , author=. Computers & Industrial Engineering , volume=. 2020 , publisher=

    2020

  31. [34]

    Procedia CIRP , volume=

    Applying natural language processing in manufacturing , author=. Procedia CIRP , volume=. 2022 , publisher=

    2022

  32. [35]

    Journal of Applied Econometrics , year=

    Manufacturing sentiment: forecasting industrial production with text analysis , author=. Journal of Applied Econometrics , year=

  33. [36]

    IEEE Transactions on Industrial Informatics , volume=

    A custom word embedding model for clustering of maintenance records , author=. IEEE Transactions on Industrial Informatics , volume=. 2021 , publisher=

    2021

  34. [37]

    Proceedings of the Design Society , volume=

    From text to images: Linking system requirements to images using joint embedding , author=. Proceedings of the Design Society , volume=. 2023 , publisher=

    2023

  35. [38]

    International Journal of Computer Integrated Manufacturing , pages=

    A predictive model based on Transformer with statistical feature embedding in manufacturing sensor dataset , author=. International Journal of Computer Integrated Manufacturing , pages=. 2025 , publisher=

    2025

  36. [39]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Is a large language model a good annotator for event extraction? , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  37. [40]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Eventgpt: Event stream understanding with multimodal large language models , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  38. [42]

    Journal of Manufacturing and Materials Processing , volume=

    Causal discovery in manufacturing: A structured literature review , author=. Journal of Manufacturing and Materials Processing , volume=. 2022 , publisher=

    2022

  39. [43]

    IEEE Transactions on Automation Science and Engineering , volume=

    Nonlinear dynamic granger causality analysis framework for root-cause diagnosis of quality-related faults in manufacturing processes , author=. IEEE Transactions on Automation Science and Engineering , volume=. 2023 , publisher=

    2023

  40. [44]

    Journal of Intelligent Manufacturing , volume=

    Data-driven dynamic causality analysis of industrial systems using interpretable machine learning and process mining , author=. Journal of Intelligent Manufacturing , volume=. 2023 , publisher=

    2023

  41. [45]

    Advanced Engineering Informatics , volume=

    An end-to-end tabular information-oriented causality event evolutionary knowledge graph for manufacturing documents , author=. Advanced Engineering Informatics , volume=. 2021 , publisher=

    2021

  42. [46]

    ACM Computing Surveys , volume=

    A Survey of Event Causality Identification: Taxonomy, Challenges, Assessment, and Prospects , author=. ACM Computing Surveys , volume=. 2025 , publisher=

    2025

  43. [47]

    Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Temporal causal modeling with graphical granger methods , author=. Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  44. [48]

    International Conference on Computer Aided Verification , pages=

    Synthesis of temporal causality , author=. International Conference on Computer Aided Verification , pages=. 2024 , organization=

    2024

  45. [50]

    Causal inference in statistics: An overview , author=

  46. [51]

    Introduction to transformers for NLP: With the hugging face library and models to solve problems , pages=

    Hugging face , author=. Introduction to transformers for NLP: With the hugging face library and models to solve problems , pages=. 2022 , publisher=

    2022

  47. [52]

    2018 IEEE/ACM 26th international symposium on quality of service (IWQoS) , pages=

    Improved adam optimizer for deep neural networks , author=. 2018 IEEE/ACM 26th international symposium on quality of service (IWQoS) , pages=. 2018 , organization=

    2018

  48. [53]

    2019 international joint conference on neural networks (IJCNN) , pages=

    A proof of local convergence for the Adam optimizer , author=. 2019 international joint conference on neural networks (IJCNN) , pages=. 2019 , organization=

    2019

  49. [55]

    Advances in neural information processing systems , volume=

    Dags with no tears: Continuous optimization for structure learning , author=. Advances in neural information processing systems , volume=

  50. [56]

    IEEE Transactions on Artificial Intelligence , year=

    Federated Learning of Dynamic Bayesian Network via Continuous Optimization from Time Series Data , author=. IEEE Transactions on Artificial Intelligence , year=

  51. [57]

    Journal of Intelligent Manufacturing , pages=

    Bayesian federated causal inference and its application in manufacturing , author=. Journal of Intelligent Manufacturing , pages=. 2025 , publisher=

    2025

  52. [59]

    Ansari, F. (2020). Cost-based text understanding to improve maintenance knowledge intelligence in manufacturing enterprises. Computers & Industrial Engineering\/ 141 , 106319

    2020

  53. [60]

    Liu, and N

    Arnold, A., Y. Liu, and N. Abe (2007). Temporal causal modeling with graphical granger methods. In Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining , pp.\ 66--75

    2007

  54. [61]

    Ban, T., L. Chen, D. Lyu, X. Wang, Q. Zhu, and H. Chen (2025). Llm-driven causal discovery via harmonized prior. IEEE Transactions on Knowledge and Data Engineering\/

    2025

  55. [62]

    Bhardwaj, A. S., A. Deep, D. Veeramani, and S. Zhou (2021). A custom word embedding model for clustering of maintenance records. IEEE Transactions on Industrial Informatics\/ 18\/ (2), 816--826

    2021

  56. [63]

    Bock, S. and M. Wei (2019). A proof of local convergence for the adam optimizer. In 2019 international joint conference on neural networks (IJCNN) , pp.\ 1--8. IEEE

    2019

  57. [64]

    Cai, R., S. Wu, J. Qiao, Z. Hao, K. Zhang, and X. Zhang (2022). Thps: Topological hawkes processes for learning causal structure on event sequences. IEEE Transactions on Neural Networks and Learning Systems\/ 35\/ (1), 479--493

    2022

  58. [65]

    Cajner, T., L. D. Crane, C. Kurz, N. Morin, P. E. Soto, and B. Vrankovich (2026). Manufacturing sentiment: forecasting industrial production with text analysis. Journal of Applied Econometrics\/

    2026

  59. [66]

    Chang, Y., D. Chen, Y. Cai, and W. Yu (2025). Anhp: Adaptive neural hawkes processes for causal structure learning on event sequences. IEEE Internet of Things Journal\/

    2025

  60. [67]

    Carroll, and B

    Chen, C., C. Carroll, and B. Morkos (2023). From text to images: Linking system requirements to images using joint embedding. Proceedings of the Design Society\/ 3 , 1985--1994

    2023

  61. [68]

    Ma, and X

    Chen, J., Y. Ma, and X. Yue (2025). Federated learning of dynamic bayesian network via continuous optimization from time series data. IEEE Transactions on Artificial Intelligence\/

    2025

  62. [69]

    Chen, R., C. Qin, W. Jiang, and D. Choi (2024). Is a large language model a good annotator for event extraction? In Proceedings of the AAAI conference on artificial intelligence , Volume 38, pp.\ 17772--17780

    2024

  63. [70]

    Yan-yi, G

    Chen, W., L. Yan-yi, G. Tie-zheng, L. Da-peng, H. Tao, L. Zhi, Y. Qing-wen, W. Hui-han, and W. Ying-you (2024). Systems engineering issues for industry applications of large language model. Applied Soft Computing\/ 151 , 111165

    2024

  64. [71]

    Cheng, Q., Z. Zeng, X. Hu, Y. Si, and Z. Liu (2025). A survey of event causality identification: Taxonomy, challenges, assessment, and prospects. ACM Computing Surveys\/ 58\/ (3), 1--37

    2025

  65. [72]

    Sun, and M

    Cui, D., Q. Sun, and M. Xie (2023). Robust statistical modeling of heterogeneity for repairable systems using multivariate gaussian convolution processes. IEEE Transactions on Reliability\/ 72\/ (4), 1493--1506

    2023

  66. [73]

    C \"u ppers, J., S. Xu, A. Musa, and J. Vreeken (2024). Causal discovery from event sequences by local cause-effect attribution. Advances in Neural Information Processing Systems\/ 37 , 24216--24241

    2024

  67. [74]

    de Oliveira, A. H., P. A. Reis, F. S. J \'u nior, M. S. Cavalcante, J. V. de Lima, L. F. Soares, and L. H. Marchiori (2024). Impact analysis of using natural language processing and large language model on automated correction of systems engineering requirements. In INCOSE International Symposium , Volume 34, pp.\ 992--1007. Wiley Online Library

    2024

  68. [75]

    Frenkel, N

    Finkbeiner, B., H. Frenkel, N. Metzger, and J. Siber (2024). Synthesis of temporal causality. In International Conference on Computer Aided Verification , pp.\ 87--111. Springer

    2024

  69. [76]

    Mei, and C

    Geigle, C., Q. Mei, and C. Zhai (2018). Feature engineering for text data. In Feature engineering for machine learning and data analytics , pp.\ 15--54. CRC Press

    2018

  70. [77]

    Zhang, D

    Gong, C., C. Zhang, D. Yao, J. Bi, W. Li, and Y. Xu (2024). Causal discovery from temporal data: An overview and new perspectives. ACM Computing Surveys\/ 57\/ (4), 1--38

    2024

  71. [78]

    Lerer, A

    Hurst, A., A. Lerer, A. P. Goucher, A. Perelman, A. Ramesh, A. Clark, A. Ostrow, A. Welihinda, A. Hayes, A. Radford, et al. (2024). Gpt-4o system card. arXiv preprint arXiv:2410.21276\/

    Pith/arXiv arXiv 2024

  72. [79]

    Jain, S. M. (2022). Hugging face. In Introduction to transformers for NLP: With the hugging face library and models to solve problems , pp.\ 51--67. Springer

    2022

  73. [80]

    Jin, Z., S. Guo, N. Chen, D. Weiskopf, D. Gotz, and N. Cao (2020). Visual causality analysis of event sequence data. IEEE transactions on visualization and computer graphics\/ 27\/ (2), 1343--1352

    2020

  74. [81]

    Kiciman, E., R. Ness, A. Sharma, and C. Tan (2023). Causal reasoning and large language models: Opening a new frontier for causality. Transactions on Machine Learning Research\/

    2023

  75. [82]

    Kwon, O.-R. and G. T. Lee (2025). A predictive model based on transformer with statistical feature embedding in manufacturing sensor dataset. International Journal of Computer Integrated Manufacturing\/ , 1--16

    2025

  76. [83]

    Li, B., X. Han, J. Liu, Y. Ding, L. Jing, Z. Zhang, J. Li, X. Du, F. Li, M. Zhang, et al. (2025). Event extraction in large language model. arXiv preprint arXiv:2512.19537\/

    arXiv 2025

  77. [84]

    Li, C. T. (2025). A poisson decomposition for information and the information-event diagram. IEEE Transactions on Information Theory\/

    2025

  78. [85]

    Li, Y., Y. Shen, Y. Nian, J. Gao, Z. Wang, C. Yu, L. Li, J. Wang, X. Hu, and Y. Zhao (2026). Mitigating hallucinations in large language models via causal reasoning. In Proceedings of the AAAI Conference on Artificial Intelligence , Volume 40, pp.\ 31852--31860

    2026

  79. [86]

    Liu, S., J. Li, G. Zhao, Y. Zhang, X. Meng, F. R. Yu, X. Ji, and M. Li (2025). Eventgpt: Event stream understanding with multimodal large language models. In Proceedings of the Computer Vision and Pattern Recognition Conference , pp.\ 29139--29149

    2025

  80. [87]

    Liu, Y., R. Cai, W. Chen, J. Qiao, Y. Yan, Z. Li, K. Zhang, and Z. Hao (2024). Tnpar: Topological neural poisson auto-regressive model for learning granger causal structure from event sequences. In Proceedings of the AAAI Conference on Artificial Intelligence , Volume 38, pp.\ 20491--20499

    2024

Showing first 80 references.