Pith. sign in

REVIEW 4 major objections 6 minor 12 cited by

RCLAgent localizes microservice root causes from a single request and reports recall@1 above the best multi-request recall@10 on the AIOps 2022 datasets.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

RCLAgent, a multi-agent recursion-of-thought system, reports Recall@1 of 71-90% on AIOps 2022 subsets from one trace, beating the Recall@10 of graph-based methods that need many requests.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection RCLAgent is a plausible multi-agent LLM system for root-cause localization, but the single-request-beats-multi-request claim is undermined by an unspecified request-selection protocol and single-run stochastic numbers. the 4 major comments →

arxiv 2508.20370 v1 pith:WS5ERANF submitted 2025-08-28 cs.SE cs.AI

Adaptive Root Cause Localization for Microservice Systems with Multi-Agent Recursion-of-Thought

Lingzhe Zhang , Tong Jia , Kangjin Wang , Weijie Hong , Chiming Duan , Minghua He , Ying Li This is my paper
classification cs.SE cs.AI
keywords root cause localizationmicroservice systemsmulti-agent LLMrecursion-of-thoughtdistributed tracingtrace analysismetrics verificationAIOps 2022
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 reading

The paper tries to settle a practical question: can an automated system find the service at fault in a microservice failure using just one abnormal request, without pre-built dependency graphs or aggregated statistics? It answers yes, with RCLAgent, a multi-agent large-language-model system modeled on how site reliability engineers actually investigate: they recursively walk the trace, expand across pod, service, and node dimensions, and verify candidates with metrics. On the AIOps 2022 datasets, the paper reports that RCLAgent's single-request recall@1 is higher than the best recall@10 of existing multi-request methods on all six subsets. If the result holds, root cause localization becomes faster and less data-hungry, and the reasoning behind the answer becomes inspectable by engineers. The paper's central contribution is the recursion-of-thought mechanism rather than a new statistical model.

Core claim

RCLAgent's central claim is that root cause localization can be reduced to a guided reasoning process over one request's trace, with metrics used only as verification, and that this process outperforms methods built on causal graphs and multi-request aggregation. The system uses data agents (Trace, Metric, Format) and thought agents (Recursion, Intermodal Inference) coordinated in three phases: initial reasoning without metrics, critical reflection that forces deeper trace traversal, and final review that consolidates the whole reasoning chain. In experiments on six AIOps 2022 subsets, the paper reports single-request recall@1 values of 71.13, 78.57, 90.24, 64.34, 65.42, and 72.41, each high

What carries the argument

The recursion-of-thought strategy is the mechanism: instead of a linear chain-of-thought, each reasoning instruction is generated from the current span, triggers a Trace Agent to fetch only that span's children, and decides whether to descend deeper, verify the candidate with the Metric Agent, or backtrace to a previously uninspected candidate. Critical reflection and final review are stage-level controls that stop premature termination. This mirrors the observed SRE pattern of recursiveness, multi-dimensional expansion, and cross-modal reasoning.

Load-bearing premise

The load-bearing premise is that the single request analyzed for each fault is chosen in a way that makes its trace sufficient to expose the true service-level root cause, and that the hand-written prompts were not tuned on the test faults; the paper does not spell out either condition.

What would settle it

Re-run the evaluation with requests selected uniformly at random from each fault window, keeping the prompts frozen before seeing the test subsets. If single-request recall@1 falls to baseline levels, the reported advantage is an artifact of request selection or prompt tuning rather than the recursion-of-thought mechanism.

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

If this is right

  • If the single-request result holds, teams could start diagnosing a fault from the first slow request instead of waiting for a window of traces to aggregate.
  • The reasoning trace is inspectable at each step, so SREs get a checkable explanation rather than an opaque ranked list.
  • Because the agent set is modular, adding new data sources such as logs or events would only require new data agents, not a new model.
  • The method is tied to the reasoning power of the underlying LLM: replacing Claude-3.5-Sonnet with smaller backbones cuts recall@1 by roughly half or more on most datasets.
  • Ablation indicates all three stages matter: initial reasoning alone reaches about 57.5% recall@1, critical reflection adds 13.3 points, and final review adds another 3.0 points on average.

Where Pith is reading between the lines

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

  • The paper does not state how the single request per fault is selected; if selection favors the most anomalous request, the single-request advantage over multi-request methods may not transfer to arbitrary requests.
  • The prompts encode SRE heuristics and were hand-crafted; a natural stress test is whether the same prompts transfer to other trace schemas and failure types without retuning, since evaluation rests on one dataset family (AIOPS 2022).
  • The backbone experiments suggest the ceiling is set by general LLM reasoning ability, so progress in smaller reasoning models could translate directly into cheaper and faster root cause localization.
  • A stronger but untested consequence is that the same recursion loop could be run incrementally as new requests arrive, reusing prior conclusions instead of restarting analysis per request.
Share X Bluesky LinkedIn Reddit HN

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

4 major / 6 minor

Summary. The paper presents RCLAgent, an LLM-based multi-agent framework for root cause localization in microservice systems. It first reports an SRE study (Section III) identifying three human diagnostic behaviors—recursiveness, multi-dimensional expansion, and cross-modal reasoning—and maps these onto data agents (Trace, Metric, Format), thought agents (Recursion, Intermodal Inference), and a three-phase coordinator (initial reasoning, critical reflection, final review). Algorithm 1 formalizes the recursion-of-thought trace walk, and Figures 6–8 provide the prompts. Evaluation on six subsets of the AIOps 2022 dataset (Section V) reports that RCLAgent's single-request Recall@1 exceeds the multi-request Recall@10 of six non-LLM baselines (Table I), outperforms mABC and CoT (Table II), and achieves high MRR under majority voting (Table III). The central claim is that a single request suffices for state-of-the-art root cause localization.

Significance. If established, the single-request result would be practically significant: it would reduce the data footprint and LLM cost of root cause analysis and demonstrate that a prompt-structured agent can outperform graph/spectrum methods that aggregate many requests. The paper's strengths include a public code repository, a clear and interpretable agent architecture, ablations isolating the three reasoning stages, and a backbone study across five LLMs. The main caveat is that the headline claim is an evaluation claim rather than a derivation, and the current evaluation omits several controls—request-selection protocol, variance reporting, and independent data—that are necessary before the comparison can be considered established.

major comments (4)
  1. [Algorithm 1 / Section V.B] The input to Algorithm 1 is 'Trace data T for a high-duration request,' but the paper never specifies how T is chosen among the anomalous requests in a failure window. Table I compares one RCLAgent run per fault against baselines that aggregate all requests. If T is selected after inspecting ground truth, or happens to be the request whose trace reaches the true service, the comparison is confounded. Please specify the request-selection protocol (e.g., first anomalous request, random request, most anomalous by entry-span latency) and report results over all requests in the window or provide a sensitivity analysis over request choices.
  2. [Tables I-IV / Section V.A.4] All RCLAgent numbers are single-run values from stochastic LLM inference (Claude-3.5-Sonnet, DeepSeek-R1-Qwen, etc.). No seeds, temperature settings, repetitions, or confidence intervals are reported. Given that the headline margins over baselines' Recall@10 are 2-9% on several datasets, sampling noise alone could change the ranking. Please repeat each condition multiple times and report means with variance/confidence intervals, or perform significance testing.
  3. [Section VI / Section V.B] The paper concedes that baseline implementations were 'fine-tuned' and 'best configurations selected' on the evaluation datasets, and all results come from one dataset family (AIOps 2022). This makes the central 'superior performance' claim fragile: the comparison mixes oracle-style baseline tuning with a single holdout family, and the abstract's 'various public datasets' overstates the evidence. Please either add a held-out or cross-validated configuration-selection procedure, or evaluate on at least one independent dataset or fault-injection benchmark.
  4. [Section V.A.4 / Section II.B] Key RCLAgent parameters are fixed without sensitivity analysis: n=3 in Equation 3, the temporal window delta in Equation 4, the 100x latency threshold that defines a 'high-duration request,' and the hand-written prompts. It is not reported whether these were selected on development sets or tuned on the test subsets. Please state the selection procedure and provide sensitivity results for n and delta; this is important for the 'adaptive' claim.
minor comments (6)
  1. [Table I] The RCLAgent column header is ambiguous: values are only Recall@1, while baselines have R1/R5/R10. Use a separate column or state clearly in the caption that RCLAgent reports single-request Recall@1.
  2. [Section V.A.2 / Tables I, III / Section VII.A] The baseline name is spelled TraceContrast in Section V.A.2, TraceConstract in Tables I and III, and TraceConstruct in Section VII.A. Please unify.
  3. [Figure 9] The y-axis is unlabeled and the x-axis labels are cut off; define CR and FR in the caption and label the axes.
  4. [Equations 7-8] Equations 7 and 8 use R and Q without precise definitions; either define them or mark them as schematic summaries rather than formal equations.
  5. [Abstract / Section V.B] The abstract says 'various public datasets,' but the evaluation uses one dataset's six subsets. Revise the wording to 'six subsets of a public dataset.'
  6. [Section V.A.2] Typo: 'We compared TraceContrast with...' should presumably be 'We compared RCLAgent with...'.

Circularity Check

0 steps flagged

No significant circularity: RCLAgent's central accuracy claim is benchmarked against external baselines, and no derivation step reduces to its inputs by construction.

full rationale

The paper's design is motivated by the authors' own qualitative SRE study (Section III), but that study is an input to design, not a prediction that is then verified by the same observation. The core evaluation (Section V.B) compares RCLAgent against external baselines on the public AIOps 2022 dataset; these are independent benchmarks, not outputs of the method. The algorithmic equations (2)-(8) are descriptive of the agent workflow and do not encode the result: Eq. (7) and (8) simply define set refinement and final formatting, and the recall/MRR metrics are standard. Self-citations appear in related work and background but are not load-bearing: no uniqueness theorem or ansatz is imported from the authors' prior papers to force the framework. The threats-to-validity section (Section VI) acknowledges that some baselines were implemented and tuned by the authors, and the paper does not specify how the single request per fault was selected for Table I (Algorithm 1 requires only 'Trace data T for a high-duration request' without describing the selection protocol). These are legitimate correctness/fairness risks that could affect the strength of the headline claim, but they are not instances of a prediction being equivalent to its inputs by definition. The recursion-of-thought strategy is a heuristic that could fail; it is not a renamed known result with the conclusion built in. Overall, the central claim has independent empirical content, so circularity is not a significant concern.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The agents (Trace, Metric, Format, Recursion, Intermodal) are software components within an existing multi-agent framework, not new postulated entities with independent evidence requirements.

free parameters (3)
  • n-sigma threshold n = 3
    Metric Agent flags a metric as anomalous iff |m(t)-mu_m| > n sigma_m (Eq. 3); n=3 is set in Sec. V.A.4 without sensitivity analysis.
  • latency anomaly threshold = 100x normal average
    A request is defined as problematic if its entry span duration exceeds 100 times the normal average latency (Sec. II.B); no justification or ablation.
  • temporal window delta = not reported
    Metric Agent evaluates metrics over [t0-delta, t0+delta] (Eq. 4); delta is never specified in the evaluation, making the metric selection procedure incompletely defined.
axioms (4)
  • domain assumption The n-sigma test with historical mean and standard deviation distinguishes root-cause-relevant metric anomalies from noise
    Used in Metric Agent (Eq. 3); no validation that this criterion selects causal metrics across the six datasets.
  • domain assumption The hand-crafted prompts and agent orchestration will make Claude-3.5-Sonnet follow the intended recursion-of-thought process consistently
    The framework is a prompting and orchestration system; its correctness depends entirely on LLM instruction following, which is not guaranteed.
  • domain assumption The undocumented SRE interviews are representative of real root cause analysis practice
    Section III bases the three design characteristics (recursiveness, multi-dimensional expansion, cross-modal reasoning) on informal insights from multiple professional SREs, with no protocol, sample size, or data.
  • domain assumption Trace and metric data in the AIOps 2022 dataset are sufficient to identify the service-level root cause from a single request
    Algorithm 1 relies on recursive trace expansion and metric verification; if root causes are only visible in aggregate statistics, the method would fail by construction.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Root Cause Localization for Microservice Systems with Multi-Agent Recursion-of-Thought." pith.science (2026). https://pith.science/paper/WS5ERANF

@misc{pith2026250820370,
  author       = {Pith},
  title        = {Pith review of: Adaptive Root Cause Localization for Microservice Systems with Multi-Agent Recursion-of-Thought},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WS5ERANF}},
  note         = {Machine review of arXiv:2508.20370}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

As contemporary microservice systems become increasingly popular and complex-often comprising hundreds or even thousands of fine-grained, interdependent subsystems-they are facing more frequent failures. Ensuring system reliability thus demands accurate root cause localization. While traces and metrics have proven to be effective data sources for this task, existing methods either heavily rely on pre-defined schemas, which struggle to adapt to evolving operational contexts, or lack interpretability in their reasoning process, thereby leaving Site Reliability Engineers (SREs) confused. In this paper, we conduct a comprehensive study on how SREs localize the root cause of failures, drawing insights from multiple professional SREs across different organizations. Our investigation reveals that human root cause analysis exhibits three key characteristics: recursiveness, multi-dimensional expansion, and cross-modal reasoning. Motivated by these findings, we introduce RCLAgent, an adaptive root cause localization method for microservice systems that leverages a multi-agent recursion-of-thought framework. RCLAgent employs a novel recursion-of-thought strategy to guide the LLM's reasoning process, effectively integrating data from multiple agents and tool-assisted analysis to accurately pinpoint the root cause. Experimental evaluations on various public datasets demonstrate that RCLAgent achieves superior performance by localizing the root cause using only a single request-outperforming state-of-the-art methods that depend on aggregating multiple requests. These results underscore the effectiveness of RCLAgent in enhancing the efficiency and precision of root cause localization in complex microservice environments.

Figures

Figures reproduced from arXiv: 2508.20370 by Chiming Duan, Kangjin Wang, Lingzhe Zhang, Minghua He, Tong Jia, Weijie Hong, Ying Li.

Figure 1
Figure 1. Figure 1: Example of Trace Log entries, known as spans, which document individual operations along the request’s execution path. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Trace Graph Example-1 For instance, we extracted a typical case from the real￾world AIOPS 2022 dataset, as shown in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Metrics Inspection of Trace Graph Example-1 [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Trace Graph Example-2 A more evident example is shown in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Architecture of RCLAgent processing functions. In this work, we primarily utilize three agents: the Trace Agent, Metrics Agent, and Format Agent. However, in practical applications, the number and types of agents can be flexibly adjusted as needed. 1) Trace Agent: Trace data is the most fundamental source for root cause localization, as it records the sequence of calls between various services. However, si… view at source ↗
Figure 6
Figure 6. Figure 6: The Prompt for Initial Reasoning [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: The Prompt for Final Review As shown in [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Ablation Experiment Overall, Initial Reasoning alone achieved an R1 (Recall@1) score of approximately 57.52%. Incorporating Critical Reflec￾tion led to an average improvement of 13.31%, while Final Review provided an additional 2.96% boost. These findings highlight the effectiveness and rationality of RCLAgent’s three-stage design. VI. THREATS TO VALIDITY Despite the promising results demonstrated by RCLAg… view at source ↗

discussion (0)

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

Forward citations

Cited by 12 Pith papers

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

  1. From Feedback Loops to Policy Updates: Reinforcement Fine-Tuning for LLM-Based Alpha Factor Discovery

    cs.CE 2026-05 unverdicted novelty 7.0

    QuantEvolver applies reinforcement fine-tuning to evolve an LLM policy for generating executable alpha factor expressions, yielding higher-quality and more complementary factors than prompt-based baselines on market b...

  2. SREGym: A Live Benchmark for AI SRE Agents with High-Fidelity Failure Scenarios

    cs.AI 2026-05 unverdicted novelty 7.0

    SREGym is a modular, open-source live benchmark with 90 high-fidelity SRE failure scenarios built on real cloud stacks for evaluating AI agents on diagnosis and mitigation tasks.

  3. Towards Robust LLM Post-Training: Automatic Failure Management for Reinforcement Fine-Tuning

    cs.SE 2026-05 unverdicted novelty 7.0

    Introduces the first benchmark for fine-grained failures in reinforcement fine-tuning of LLMs and an automatic management framework that detects, diagnoses, and remediates them.

  4. E2E-REME: Towards End-to-End Microservices Auto-Remediation via Experience-Simulation Reinforcement Fine-Tuning

    cs.SE 2026-04 unverdicted novelty 7.0

    E2E-REME outperforms nine LLMs in accuracy and efficiency for end-to-end microservice remediation by using experience-simulation reinforcement fine-tuning on a new benchmark called MicroRemed.

  5. LLM4Log: A Systematic Review of Large Language Model-based Log Analysis

    cs.SE 2026-03 accept novelty 7.0

    LLM4Log is a systematic review of 145 papers on LLM-based log analysis that delivers a unified taxonomy, design patterns, and open challenges for reliable adoption in AIOps.

  6. Bifrost: Empowering Pretrained Language Model with Fallibility Representation for Log-Based Fault Diagnosis

    cs.SE 2026-07 conditional novelty 6.0

    Bifrost fine-tunes BART-Base on logs with three self-supervised contrastive tasks, beating general PLMs by roughly 10-20% on anomaly detection, root-cause localization, and fault identification.

  7. TopoEvo: A Topology-Aware Self-Evolving Multi-Agent Framework for Root Cause Analysis in Microservices

    cs.AI 2026-05 unverdicted novelty 6.0

    TopoEvo is a topology-aware self-evolving multi-agent framework for root cause analysis in microservices that uses multimodal alignment, vector-quantized symptom tokens, and a hypothesis-evidence-test workflow to sepa...

  8. Towards In-Depth Root Cause Localization for Microservices with Multi-Agent Recursion-of-Thought

    cs.SE 2026-05 unverdicted novelty 6.0

    RCLAgent uses multi-agent recursion-of-thought with parallel reasoning on trace graphs to outperform prior methods in root cause localization accuracy and efficiency for microservice systems.

  9. SREGym: A Live Benchmark for AI SRE Agents with High-Fidelity Failure Scenarios

    cs.AI 2026-05 conditional novelty 6.0

    SREGym is an open-source benchmark of 90 live cloud failures for AI SRE agents, revealing up to 40-percentage-point differences in agent success across failure types.

  10. Anomaly Detection and Root Cause Analysis for Microservice Systems

    cs.SE 2026-06 unverdicted novelty 5.0

    Thesis proposes BARO for metrics, EventADL for events, TORAI for multimodal RCA without call graphs, and RCAEval benchmark with systematic evaluation of causal methods.

  11. SREGym: A Live Benchmark for AI SRE Agents with High-Fidelity Failure Scenarios

    cs.AI 2026-05 unverdicted novelty 5.0

    SREGym supplies 90 high-fidelity SRE tasks in a live environment to measure how well frontier AI agents handle diverse faults, noises, and complex failure modes such as metastable and correlated failures.

  12. LLM4Log: A Systematic Review of Large Language Model-based Log Analysis

    cs.SE 2026-03 unverdicted novelty 4.0

    Systematic review of 145 papers on LLM-based log analysis, providing a unified taxonomy, common design patterns, evaluation practices, and challenges for deployment under drift and limited labels.

Reference graph

Works this paper leans on

65 extracted references · 48 canonical work pages · cited by 9 Pith papers

  1. [1]

    Fault analysis and debugging of microservice systems: Industrial survey, benchmark system, and empirical study,

    X. Zhou, X. Peng, T. Xie, J. Sun, C. Ji, W. Li, and D. Ding, “Fault analysis and debugging of microservice systems: Industrial survey, benchmark system, and empirical study,”IEEE Transactions on Software Engineering, vol. 47, no. 2, pp. 243–260, 2018

    2018

  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

    Pith/arXiv arXiv 2024

  3. [3]

    A survey of aiops in the era of large language models,

    L. Zhang, T. Jia, M. Jia, Y . Wu, A. Liu, Y . Yang, Z. Wu, X. Hu, P. Yu, and Y . Li, “A survey of aiops in the era of large language models,”ACM Computing Surveys, 2025

    2025

  4. [4]

    Developing self-adaptive microservice systems: Challenges and directions,

    N. C. Mendonc ¸a, P. Jamshidi, D. Garlan, and C. Pahl, “Developing self-adaptive microservice systems: Challenges and directions,” IEEE Software, vol. 38, no. 2, pp. 70–79, 2019

    work page 2019

  5. [5]

    Design, monitoring, and testing of microservices systems: The practitioners’ perspective,

    M. Waseem, P. Liang, M. Shahin, A. Di Salle, and G. M ´arquez, “Design, monitoring, and testing of microservices systems: The practitioners’ perspective,” Journal of Systems and Software , vol. 182, p. 111061, 2021

    work page 2021

  6. [6]

    Towards close-to-zero runtime collection overhead: Raft-based anomaly diagnosis on system faults for distributed storage system,

    L. Zhang, T. Jia, M. Jia, H. Liu, Y . Yang, Z. Wu, and Y . Li, “Towards close-to-zero runtime collection overhead: Raft-based anomaly diagnosis on system faults for distributed storage system,” IEEE Transactions on Services Computing, 2024

    2024

  7. [7]

    Time-tired compaction: An elastic compaction scheme for lsm-tree based time-series database,

    L.-Z. Zhang, X.-D. Huang, Y .-K. Wang, J.-L. Qiao, S.-X. Song, and J.- M. Wang, “Time-tired compaction: An elastic compaction scheme for lsm-tree based time-series database,” Advanced Engineering Informatics, vol. 59, p. 102224, 2024

    work page 2024

  8. [8]

    Separation or not: On handing out-of-order time-series data in leveled lsm-tree,

    Y . Kang, X. Huang, S. Song, L. Zhang, J. Qiao, C. Wang, J. Wang, and J. Feinauer, “Separation or not: On handing out-of-order time-series data in leveled lsm-tree,” in 2022 IEEE 38th International Conference on Data Engineering (ICDE) . IEEE, 2022, pp. 3340–3352

    work page 2022

  9. [9]

    Multivariate log- based anomaly detection for distributed database,

    L. Zhang, T. Jia, M. Jia, Y . Li, Y . Yang, and Z. Wu, “Multivariate log- based anomaly detection for distributed database,” in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, pp. 4256–4267

    2024

  10. [10]

    Reducing events to augment log-based anomaly detection models: An empirical study,

    L. Zhang, T. Jia, K. Wang, M. Jia, Y . Yang, and Y . Li, “Reducing events to augment log-based anomaly detection models: An empirical study,” in Proceedings of the 18th ACM/IEEE International Symposium on Empirical Software Engineering and Measurement , 2024, pp. 538– 548

    work page 2024

  11. [11]

    Inter- dependent causal networks for root cause localization,

    D. Wang, Z. Chen, J. Ni, L. Tong, Z. Wang, Y . Fu, and H. Chen, “Inter- dependent causal networks for root cause localization,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023, pp. 5051–5060

    work page 2023

  12. [12]

    Failure diagnosis in microservice systems: A comprehensive survey and analysis,

    S. Zhang, S. Xia, W. Fan, B. Shi, X. Xiong, Z. Zhong, M. Ma, Y . Sun, and D. Pei, “Failure diagnosis in microservice systems: A comprehensive survey and analysis,” ACM Transactions on Software Engineering and Methodology, 2024

    work page 2024

  13. [13]

    A survey on intelligent management of alerts and incidents in it services,

    Q. Yu, N. Zhao, M. Li, Z. Li, H. Wang, W. Zhang, K. Sui, and D. Pei, “A survey on intelligent management of alerts and incidents in it services,” Journal of Network and Computer Applications , p. 103842, 2024

    work page 2024

  14. [14]

    Interpretable failure localization for microservice systems based on graph autoencoder,

    Y . Sun, Z. Lin, B. Shi, S. Zhang, S. Ma, P. Jin, Z. Zhong, L. Pan, Y . Guo, and D. Pei, “Interpretable failure localization for microservice systems based on graph autoencoder,” ACM Transactions on Software Engineering and Methodology , vol. 34, no. 2, pp. 1–28, 2025

    work page 2025

  15. [15]

    Hemirca: Fine-grained root cause analysis for microservices with heterogeneous data sources,

    Z. Zhu, C. Lee, X. Tang, and P. He, “Hemirca: Fine-grained root cause analysis for microservices with heterogeneous data sources,” ACM Transactions on Software Engineering and Methodology , vol. 33, no. 8, pp. 1–25, 2024

    work page 2024

  16. [16]

    Microservice root cause analysis with limited observability through intervention recognition in the latent space,

    Z. Xie, S. Zhang, Y . Geng, Y . Zhang, M. Ma, X. Nie, Z. Yao, L. Xu, Y . Sun, W. Li et al. , “Microservice root cause analysis with limited observability through intervention recognition in the latent space,” in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2024, pp. 6049–6060

    work page 2024

  17. [17]

    Kgroot: A knowledge graph-enhanced method for root cause analysis,

    T. Wang, G. Qi, and T. Wu, “Kgroot: A knowledge graph-enhanced method for root cause analysis,” Expert Systems with Applications , vol. 255, p. 124679, 2024

    work page 2024

  18. [18]

    E-log: Fine-grained elastic log-based anomaly detection and diagnosis for databases,

    L. Zhang, T. Jia, X. Tan, X. Huang, M. Jia, H. Liu, Z. Wu, and Y . Li, “E-log: Fine-grained elastic log-based anomaly detection and diagnosis for databases,” IEEE Transactions on Services Computing , 2025

    work page 2025

  19. [19]

    Microscope: Pinpoint performance issues with causal graphs in micro-service environments,

    J. Lin, P. Chen, and Z. Zheng, “Microscope: Pinpoint performance issues with causal graphs in micro-service environments,” in Service-Oriented Computing: 16th International Conference, ICSOC 2018, Hangzhou, China, November 12-15, 2018, Proceedings 16 . Springer, 2018, pp. 3–20

    work page 2018

  20. [20]

    Causal inference-based root cause analysis for online service systems with intervention recognition,

    M. Li, Z. Li, K. Yin, X. Nie, W. Zhang, K. Sui, and D. Pei, “Causal inference-based root cause analysis for online service systems with intervention recognition,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 3230– 3240

    work page 2022

  21. [21]

    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

    work page 2024

  22. [22]

    Microrank: End-to-end latency issue localization with extended spectrum analysis in microservice environments,

    G. Yu, P. Chen, H. Chen, Z. Guan, Z. Huang, L. Jing, T. Weng, X. Sun, and X. Li, “Microrank: End-to-end latency issue localization with extended spectrum analysis in microservice environments,” in Proceedings of the Web Conference 2021 , 2021, pp. 3087–3098

    work page 2021

  23. [23]

    Tracerank: Abnormal service localization with dis-aggregated end-to-end tracing data in cloud native systems,

    G. Yu, Z. Huang, and P. Chen, “Tracerank: Abnormal service localization with dis-aggregated end-to-end tracing data in cloud native systems,” Journal of Software: Evolution and Process , vol. 35, no. 10, p. e2413, 2023

    work page 2023

  24. [24]

    {CRISP}: Critical path analysis of {Large-Scale} mi- croservice architectures,

    Z. Zhang, M. K. Ramanathan, P. Raj, A. Parwal, T. Sherwood, and M. Chabbi, “ {CRISP}: Critical path analysis of {Large-Scale} mi- croservice architectures,” in 2022 USENIX Annual Technical Conference (USENIX ATC 22) , 2022, pp. 655–672

    work page 2022

  25. [25]

    Trace-based multi-dimensional root cause localization of performance issues in mi- croservice systems,

    C. Zhang, Z. Dong, X. Peng, B. Zhang, and M. Chen, “Trace-based multi-dimensional root cause localization of performance issues in mi- croservice systems,” in Proceedings of the IEEE/ACM 46th International Conference on Software Engineering , 2024, pp. 1–12

    work page 2024

  26. [26]

    mabc: Multi-agent blockchain-inspired collaboration for root cause analysis in micro-services architecture,

    W. Zhang, H. Guo, J. Yang, Z. Tian, Y . Zhang, Y . Chaoran, Z. Li, T. Li, X. Shi, L. Zheng et al. , “mabc: Multi-agent blockchain-inspired collaboration for root cause analysis in micro-services architecture,” in Findings of the Association for Computational Linguistics: EMNLP 2024, 2024, pp. 4017–4033

    work page 2024

  27. [27]

    Rcagent: Cloud root cause analysis by autonomous agents with tool-augmented large language models,

    Z. Wang, Z. Liu, Y . Zhang, A. Zhong, J. Wang, F. Yin, L. Fan, L. Wu, and Q. Wen, “Rcagent: Cloud root cause analysis by autonomous agents with tool-augmented large language models,” in Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024, pp. 4966–4974

    work page 2024

  28. [28]

    Characterizing job microarchitectural profiles at scale: Dataset and analysis,

    K. Wang, Y . Li, C. Wang, T. Jia, K. Chow, Y . Wen, Y . Dou, G. Xu, C. Hou, J. Yao et al., “Characterizing job microarchitectural profiles at scale: Dataset and analysis,” in Proceedings of the 51st International Conference on Parallel Processing, 2022, pp. 1–11

    work page 2022

  29. [29]

    Capturing request execution path for understanding service behavior and detecting anomalies without code instrumentation,

    Y . Yang, L. Wang, J. Gu, and Y . Li, “Capturing request execution path for understanding service behavior and detecting anomalies without code instrumentation,” IEEE Transactions on Services Computing , vol. 16, no. 2, pp. 996–1010, 2022

    work page 2022

  30. [30]

    Network-centric distributed tracing with deepflow: Troubleshooting your microservices in zero code,

    J. Shen, H. Zhang, Y . Xiang, X. Shi, X. Li, Y . Shen, Z. Zhang, Y . Wu, X. Yin, J. Wang et al. , “Network-centric distributed tracing with deepflow: Troubleshooting your microservices in zero code,” in Proceedings of the ACM SIGCOMM 2023 Conference , 2023, pp. 420– 437

    work page 2023

  31. [31]

    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. Zhou et 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

  32. [32]

    Enjoy your observability: an industrial survey of microservice tracing and analysis,

    B. Li, X. Peng, Q. Xiang, H. Wang, T. Xie, J. Sun, and X. Liu, “Enjoy your observability: an industrial survey of microservice tracing and analysis,” Empirical Software Engineering , vol. 27, pp. 1–28, 2022

    work page 2022

  33. [33]

    Characterizing microservice dependency and performance: Alibaba trace analysis,

    S. Luo, H. Xu, C. Lu, K. Ye, G. Xu, L. Zhang, Y . Ding, J. He, and C. Xu, “Characterizing microservice dependency and performance: Alibaba trace analysis,” in Proceedings of the ACM symposium on cloud computing, 2021, pp. 412–426

    work page 2021

  34. [34]

    X. Zhou, X. Peng, T. Xie, J. Sun, C. Ji, D. Liu, Q. Xiang, and C. He, “Latent error prediction and fault localization for microservice ADAPTIVE ROOT CAUSE LOCALIZATION FOR MICROSERVICE SYSTEMS WITH MULTI-AGENT RECURSION-OF-THOUGHT 12 applications by learning from system trace logs,” in Proceedings of the 2019 27th ACM joint meeting on European software en...

    work page 2019

  35. [35]

    Seer: Leveraging big data to navigate the complexity of performance debugging in cloud microservices,

    Y . Gan, Y . Zhang, K. Hu, D. Cheng, Y . He, M. Pancholi, and C. Delimitrou, “Seer: Leveraging big data to navigate the complexity of performance debugging in cloud microservices,” in Proceedings of the twenty-fourth international conference on architectural support for programming languages and operating systems , 2019, pp. 19–33

    work page 2019

  36. [36]

    Unsupervised detection of microservice trace anomalies through service-level deep bayesian networks,

    P. Liu, H. Xu, Q. Ouyang, R. Jiao, Z. Chen, S. Zhang, J. Yang, L. Mo, J. Zeng, W. Xue et al., “Unsupervised detection of microservice trace anomalies through service-level deep bayesian networks,” in 2020 IEEE 31st International Symposium on Software Reliability Engineering (ISSRE). IEEE, 2020, pp. 48–58

    2020

  37. [37]

    Sage: practical and scalable ml-driven performance debugging in microservices,

    Y . Gan, M. Liang, S. Dev, D. Lo, and C. Delimitrou, “Sage: practical and scalable ml-driven performance debugging in microservices,” in Proceedings of the 26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems , 2021, pp. 135–151

    work page 2021

  38. [38]

    Practical root cause localization for microservice systems via trace analysis,

    Z. Li, J. Chen, R. Jiao, N. Zhao, Z. Wang, S. Zhang, Y . Wu, L. Jiang, L. Yan, Z. Wanget al., “Practical root cause localization for microservice systems via trace analysis,” in 2021 IEEE/ACM 29th International Symposium on Quality of Service (IWQOS) . IEEE, 2021, pp. 1–10

    work page 2021

  39. [39]

    Lag-llama: Towards foundation models for time series forecasting,

    K. Rasul, A. Ashok, A. R. Williams, A. Khorasani, G. Adamopoulos, R. Bhagwatkar, M. Bilo ˇs, H. Ghonia, N. Hassen, A. Schneider et al. , “Lag-llama: Towards foundation models for time series forecasting,” in R0-FoMo: Robustness of Few-shot and Zero-shot Learning in Large Foundation Models, 2023

    work page 2023

  40. [40]

    Timer: generative pre-trained transformers are large time series models,

    Y . Liu, H. Zhang, C. Li, X. Huang, J. Wang, and M. Long, “Timer: generative pre-trained transformers are large time series models,” in Proceedings of the 41st International Conference on Machine Learning , 2024, pp. 32 369–32 399

    work page 2024

  41. [41]

    A decoder-only foundation model for time-series forecasting,

    A. Das, W. Kong, R. Sen, and Y . Zhou, “A decoder-only foundation model for time-series forecasting,” in Forty-first International Confer- ence on Machine Learning , 2024

    2024

  42. [42]

    Shellgpt: Gener- ative pre-trained transformer model for shell language understanding,

    J. Shi, S. Jiang, B. Xu, J. Liang, Y . Xiao, and W. Wang, “Shellgpt: Gener- ative pre-trained transformer model for shell language understanding,” in 2023 IEEE 34th International Symposium on Software Reliability Engineering (ISSRE). IEEE, 2023, pp. 671–682

    2023

  43. [43]

    Anomalyllm: Few-shot anomaly edge detection for dynamic graphs using large language models,

    S. Liu, D. Yao, L. Fang, Z. Li, W. Li, K. Feng, X. Ji, and J. Bi, “Anomalyllm: Few-shot anomaly edge detection for dynamic graphs using large language models,” arXiv preprint arXiv:2405.07626 , 2024

    Pith/arXiv arXiv 2024

  44. [44]

    Unitime: A language-empowered unified model for cross-domain time series forecasting,

    X. Liu, J. Hu, Y . Li, S. Diao, Y . Liang, B. Hooi, and R. Zimmermann, “Unitime: A language-empowered unified model for cross-domain time series forecasting,” in Proceedings of the ACM Web Conference 2024 , 2024, pp. 4095–4106

    work page 2024

  45. [45]

    Owl: A large language model for it operations,

    H. Guo, J. Yang, J. Liu, L. Yang, L. Chai, J. Bai, J. Peng, X. Hu, C. Chen, D. Zhang et al., “Owl: A large language model for it operations,” arXiv preprint arXiv:2309.09298, 2023

    Pith/arXiv arXiv 2023

  46. [46]

    Loglm: From task-based to instruction-based automated log analysis,

    Y . Liu, Y . Ji, S. Tao, M. He, W. Meng, S. Zhang, Y . Sun, Y . Xie, B. Chen, and H. Yang, “Loglm: From task-based to instruction-based automated log analysis,” arXiv preprint arXiv:2410.09352 , 2024

    Pith/arXiv arXiv 2024

  47. [47]

    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. Wen et al., “Automatic root cause analysis via large language models for cloud incidents,” in Proceedings of the Nineteenth European Conference on Computer Systems , 2024, pp. 674–688

    work page 2024

  48. [48]

    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. Lin et al., “Xpert: Empowering incident management with query recommendations via large language models,” in Proceedings of the IEEE/ACM 46th International Conference on Software Engineer- ing, 2024, pp. 1–13

    work page 2024

  49. [49]

    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,” in Companion Proceedings of the 32nd ACM International Conference on the Foundations of Software Engineering, 2024, pp. 388–398

    2024

  50. [50]

    A holistic view of ai-driven network incident management,

    P. Hamadanian, B. Arzani, S. Fouladi, S. K. R. Kakarla, R. Fonseca, D. Billor, A. Cheema, E. Nkposong, and R. Chandra, “A holistic view of ai-driven network incident management,” in Proceedings of the 22nd ACM Workshop on Hot Topics in Networks , 2023, pp. 180–188

    work page 2023

  51. [51]

    Raglog: Log anomaly detection using retrieval augmented generation,

    J. Pan, W. S. Liang, and Y . Yidi, “Raglog: Log anomaly detection using retrieval augmented generation,” in 2024 IEEE World Forum on Public Safety Technology (WFPST). IEEE, 2024, pp. 169–174

    work page 2024

  52. [52]

    Lograg: Semi-supervised log-based anomaly detection with retrieval- augmented generation,

    W. Zhang, Q. Zhang, E. Yu, Y . Ren, Y . Meng, M. Qiu, and J. Wang, “Lograg: Semi-supervised log-based anomaly detection with retrieval- augmented generation,” in 2024 IEEE International Conference on Web Services (ICWS). IEEE, 2024, pp. 1100–1102

    work page 2024

  53. [53]

    Xraglog: A resource- efficient and context-aware log-based anomaly detection method using retrieval-augmented generation,

    L. Zhang, T. Jia, M. Jia, Y . Wu, H. Liu, and Y . Li, “Xraglog: A resource- efficient and context-aware log-based anomaly detection method using retrieval-augmented generation,” in AAAI 2025 Workshop on Preventing and Detecting LLM Misinformation (PDLM) , 2025

    work page 2025

  54. [54]

    Scalalog: Scalable log-based failure diagnosis using llm,

    ——, “Scalalog: Scalable log-based failure diagnosis using llm,” in ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2025, pp. 1–5

    work page 2025

  55. [55]

    Agentfm: Role-aware failure management for distributed databases with llm-driven multi-agents,

    L. Zhang, Y . Zhai, T. Jia, X. Huang, C. Duan, and Y . Li, “Agentfm: Role-aware failure management for distributed databases with llm-driven multi-agents,” arXiv preprint arXiv:2504.06614 , 2025

    Pith/arXiv arXiv 2025

  56. [56]

    Thinkfl: Self-refining failure localization for microservice sys- tems via reinforcement fine-tuning,

    L. Zhang, Y . Zhai, T. Jia, C. Duan, S. Yu, J. Gao, B. Ding, Z. Wu, and Y . Li, “Thinkfl: Self-refining failure localization for microservice sys- tems via reinforcement fine-tuning,” arXiv preprint arXiv:2504.18776 , 2025

    arXiv 2025

  57. [57]

    Logdb: Multivariate log-based failure diagnosis for distributed databases (extended from multilog),

    L. Zhang, T. Jia, M. Jia, and Y . Li, “Logdb: Multivariate log-based failure diagnosis for distributed databases (extended from multilog),” arXiv preprint arXiv:2505.01676 , 2025

    Pith/arXiv arXiv 2025

  58. [58]

    Eagerlog: Active learning enhanced retrieval augmented gen- eration for log-based anomaly detection,

    C. Duan, T. Jia, Y . Yang, G. Liu, J. Liu, H. Zhang, Q. Zhou, Y . Li, and G. Huang, “Eagerlog: Active learning enhanced retrieval augmented gen- eration for log-based anomaly detection,” in ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    work page 2025

  59. [59]

    Weakly- supervised log-based anomaly detection with inexact labels via multi- instance learning,

    M. He, T. Jia, C. Duan, H. Cai, Y . Li, and G. Huang, “Weakly- supervised log-based anomaly detection with inexact labels via multi- instance learning,” in 2025 IEEE/ACM 47th International Conference on Software Engineering (ICSE) . IEEE Computer Society, 2025, pp. 726–726

    work page 2025

  60. [60]

    Famos: Fault diagnosis for microservice systems through ef- fective 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 ef- fective multi-modal data fusion,” in 2025 IEEE/ACM 47th International Conference on Software Engineering (ICSE) . IEEE Computer Society, 2025, pp. 610–610

    work page 2025

  61. [61]

    Logcae: An approach for log-based anomaly detection with active learning and contrastive learning,

    P. Xiao, T. Jia, C. Duan, H. Cai, Y . Li, and G. Huang, “Logcae: An approach for log-based anomaly detection with active learning and contrastive learning,” in 2024 IEEE 35th International Symposium on Software Reliability Engineering (ISSRE) . IEEE, 2024, pp. 144–155

    work page 2024

  62. [62]

    Llmelog: An approach for anomaly detection based on llm-enriched log events,

    M. He, T. Jia, C. Duan, H. Cai, Y . Li, and G. Huang, “Llmelog: An approach for anomaly detection based on llm-enriched log events,” in 2024 IEEE 35th International Symposium on Software Reliability Engineering (ISSRE). IEEE, 2024, pp. 132–143

    work page 2024

  63. [63]

    Afalog: A general augmentation framework for log-based anomaly detection with active learning,

    C. Duan, T. Jia, H. Cai, Y . Li, and G. Huang, “Afalog: A general augmentation framework for log-based anomaly detection with active learning,” in 2023 IEEE 34th International Symposium on Software Reliability Engineering (ISSRE) . IEEE, 2023, pp. 46–56

    work page 2023

  64. [64]

    Aclog: An approach to detecting anomalies from system logs with active learning,

    C. Duan, T. Jia, Y . Li, and G. Huang, “Aclog: An approach to detecting anomalies from system logs with active learning,” in 2023 IEEE International Conference on Web Services (ICWS) . IEEE, 2023, pp. 436–443

    work page 2023

  65. [65]

    A survey on parallel text generation: From parallel decoding to diffusion language models,

    L. Zhang, L. Fang, C. Duan, M. He, L. Pan, P. Xiao, S. Huang, Y . Zhai, X. Hu, P. S. Yu et al. , “A survey on parallel text generation: From parallel decoding to diffusion language models,” arXiv preprint arXiv:2508.08712, 2025

    arXiv 2025

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.