pith. sign in

arxiv: 2606.04435 · v1 · pith:L5F43UO5new · submitted 2026-06-03 · 💻 cs.AI · cs.CL· cs.CR· cs.IR

Cascading Hallucination in Agentic RAG: The CHARM Framework for Detection and Mitigation

Saroj Mishra This is my paper

Pith reviewed 2026-06-28 06:30 UTC · model grok-4.3

classification 💻 cs.AI cs.CLcs.CRcs.IR
keywords cascading hallucinationagentic RAGhallucination detectionerror propagationmulti-step reasoningfact verificationconsistency trackingCHARM framework
0
0 comments X

The pith

The CHARM framework detects cascading hallucinations in multi-step agentic RAG systems by adding stage-level verification and consistency tracking.

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

Agentic RAG pipelines for complex tasks often fail when an early hallucination spreads through later steps, creating final answers that look confident but are wrong. Existing detectors that only check the final output miss most of these cascades. CHARM formalizes the problem with a taxonomy of four cascade patterns and adds four components that run alongside the pipeline to catch errors early. In tests on multi-hop QA datasets, it detects 89.4 percent of cascades, cuts error spread by 82.1 percent, and adds little extra time per step.

Core claim

Cascading hallucination is a distinct failure mode in agentic RAG where errors at early stages propagate and amplify across successive reasoning steps. CHARM provides four components—stage-level fact verification, cross-stage consistency tracking, confidence propagation monitoring, and cascade resolution triggering—that detect and interrupt this propagation without replacing the underlying pipeline, yielding an 89.4% cascade detection rate and 82.1% error propagation reduction.

What carries the argument

CHARM, an architectural framework consisting of stage-level fact verification, cross-stage consistency tracking, confidence propagation monitoring, and cascade resolution triggering that operates alongside standard agentic RAG pipelines.

If this is right

  • Stage-level fact verification identifies errors before they reach later reasoning steps.
  • Cross-stage consistency tracking reveals when intermediate outputs conflict with prior steps.
  • Confidence propagation monitoring flags decreasing reliability across the pipeline.
  • Cascade resolution triggering interrupts error spread before the final output is generated.
  • The framework integrates with existing human-in-the-loop oversight for production use.

Where Pith is reading between the lines

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

  • Similar stage-wise checks could be adapted to other multi-step agent architectures like planning or tool-use systems.
  • Evaluating CHARM on real-world production pipelines rather than only benchmarks would test its practical robustness.
  • The taxonomy of cascade patterns may help classify failures in related systems such as chain-of-thought reasoning.
  • Reducing error propagation this way could improve reliability in safety-critical applications of agentic AI.

Load-bearing premise

Stage-level fact verification and cross-stage consistency tracking can be implemented accurately on retrieved content without introducing new errors or requiring changes to the underlying agentic pipeline architecture.

What would settle it

Running CHARM on a multi-hop QA dataset where a known early-stage hallucination propagates to the final answer but CHARM reports no cascade would falsify the detection effectiveness.

Figures

Figures reproduced from arXiv: 2606.04435 by Saroj Mishra.

Figure 1
Figure 1. Figure 1: A standard 5-stage agentic RAG pipeline. The context output is continuously passed forward as the definitive input for subsequent reasoning stages, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: DAG-based representation G = (V, E) of a multi-step agentic pipeline. The highlighted path demonstrates a cascading hallucination where the cumulative error propagation weight P(ϵi+1|ϵi) forces a high terminal divergence from ground truth. D. Four-Type Cascading Hallucination Taxonomy Because errors enter the DAG at different nodes and com￾pound in different ways, generalized detection is insufficient. We … view at source ↗
Figure 3
Figure 3. Figure 3: The CHARM System Architecture. The standard agentic pipeline (left) executes normally while the parallel CHARM layer (center) monitors inter-stage [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
read the original abstract

Multi-step agentic retrieval-augmented generation (RAG) pipelines have demonstrated significant capability for complex reasoning tasks, yet remain vulnerable to a class of failure that existing hallucination detection mechanisms systematically miss: cascading hallucination, where errors introduced at early pipeline stages propagate and amplify across successive reasoning steps, producing confident but factually incorrect final outputs. To address this vulnerability, we formalize cascading hallucination as a distinct failure mode in agentic RAG systems, present a four-type taxonomy of cascade patterns, and introduce CHARM (Cascading Hallucination Aware Resolution and Mitigation), an architectural framework for detecting and interrupting error propagation in multi-step reasoning pipelines. CHARM comprises four components - stage-level fact verification, cross-stage consistency tracking, confidence propagation monitoring, and cascade resolution triggering - that operate alongside standard agentic RAG pipelines without requiring architectural replacement. We evaluate CHARM on HotpotQA, MuSiQue, 2WikiMultiHopQA, and a custom adversarial dataset across LangChain agentic pipeline configurations, achieving an 89.4% cascade detection rate with a 5.3% false positive rate and 215 ms +/- 18 ms average latency overhead per stage, achieving an error propagation reduction of 82.1%, compared to 18.5% for output-level detectors. Component ablations confirm that each detection module contributes meaningfully to overall cascade coverage. CHARM integrates with human-in-the-loop oversight frameworks to provide a complete reliability and governance stack for production agentic AI deployment.

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

1 major / 2 minor

Summary. The manuscript formalizes 'cascading hallucination' as a distinct failure mode in multi-step agentic RAG pipelines where early-stage errors propagate and amplify across reasoning steps. It presents a four-type taxonomy of cascade patterns and introduces the CHARM framework with four components (stage-level fact verification, cross-stage consistency tracking, confidence propagation monitoring, and cascade resolution triggering) that can be added to existing pipelines without replacement. Evaluation on HotpotQA, MuSiQue, 2WikiMultiHopQA, and a custom adversarial dataset across LangChain configurations reports 89.4% cascade detection rate, 5.3% false positive rate, 215 ms +/- 18 ms latency overhead per stage, and 82.1% error propagation reduction (vs. 18.5% for output-level detectors), with component ablations confirming contributions; integration with human-in-the-loop oversight is also discussed.

Significance. If the empirical results hold and the verification components prove reliable on multi-hop content, this would represent a practical contribution to robustness in agentic RAG by targeting error propagation specifically, with low overhead and compatibility with existing systems and governance frameworks.

major comments (1)
  1. [Evaluation (as described in abstract)] Evaluation (abstract): The central claims of 89.4% cascade detection and 82.1% error propagation reduction rest on the untested premise that stage-level fact verification and cross-stage consistency tracking achieve high accuracy on the same retrieved passages used by the agentic pipeline. On HotpotQA and MuSiQue, retrieved content often contains partial, conflicting, or ambiguous facts across hops; no independent validation, error analysis, or ablation isolating verifier accuracy is described to rule out missed cascades or spurious triggers.
minor comments (2)
  1. [Abstract] The custom adversarial dataset is referenced but its construction, size, adversarial strategy, and relation to the standard benchmarks are not described.
  2. [Abstract] Specific LangChain agentic pipeline configurations (e.g., number of stages, retrieval settings) used for the reported metrics are not detailed.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on our evaluation methodology. We address the concern point by point below.

read point-by-point responses

  1. Referee: The central claims of 89.4% cascade detection and 82.1% error propagation reduction rest on the untested premise that stage-level fact verification and cross-stage consistency tracking achieve high accuracy on the same retrieved passages used by the agentic pipeline. On HotpotQA and MuSiQue, retrieved content often contains partial, conflicting, or ambiguous facts across hops; no independent validation, error analysis, or ablation isolating verifier accuracy is described to rule out missed cascades or spurious triggers.

    Authors: We acknowledge that the manuscript does not include an independent validation or error analysis isolating the accuracy of the stage-level fact verifier on the retrieved passages themselves. The reported 89.4% detection rate and component ablations reflect end-to-end cascade coverage against ground-truth annotations rather than direct verifier precision/recall on ambiguous or conflicting passages. We agree this leaves open the possibility of missed cascades or spurious triggers and will add a dedicated verifier accuracy analysis (including precision/recall on HotpotQA and MuSiQue passages) plus an ablation isolating verifier performance in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on external benchmarks

full rationale

The paper introduces the CHARM framework with four components and reports empirical results (89.4% detection rate, 82.1% error reduction) on public benchmarks HotpotQA, MuSiQue, 2WikiMultiHopQA plus a custom dataset. No equations, derivations, or self-citations are shown that reduce any prediction or result to fitted inputs or self-definitions by construction. Component ablations and latency measurements are presented as direct experimental outcomes rather than tautological outputs. This is the normal case of an empirical systems paper with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

Review performed on abstract only; full text unavailable so ledger reflects only claims visible in the abstract. No free parameters are named. The taxonomy and components are presented as novel contributions without external grounding.

axioms (1)
  • domain assumption Standard multi-hop QA benchmarks accurately represent cascading hallucination behavior in production agentic systems
    Evaluation and claims rest on results from HotpotQA, MuSiQue, 2WikiMultiHopQA and a custom adversarial set.
invented entities (2)
  • Cascading hallucination no independent evidence
    purpose: To name and taxonomize the error-propagation failure mode
    Newly formalized in the paper as distinct from existing hallucination categories.
  • CHARM framework (four components) no independent evidence
    purpose: Stage-level detection and mitigation architecture
    Newly introduced architectural framework.

pith-pipeline@v0.9.1-grok · 5807 in / 1543 out tokens · 57925 ms · 2026-06-28T06:30:49.524827+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

45 extracted references · 17 canonical work pages · 7 internal anchors

  1. [1]

    Sok: Agentic retrieval-augmented generation (rag): Taxon- omy, architectures, evaluation, and research directions,

    S. Mishra, S. Niroula, U. Yadav, D. Thakur, S. Gyawali, and S. Gaire, “Sok: Agentic retrieval-augmented generation (rag): Taxon- omy, architectures, evaluation, and research directions,”arXiv preprint arXiv:2603.07379, 2026

    work page arXiv 2026

  2. [2]

    Artificial intelligence risk management framework: Generative ai profile (nist ai 600-1),

    National Institute of Standards and Technology, “Artificial intelligence risk management framework: Generative ai profile (nist ai 600-1),” U.S. Department of Commerce, Tech. Rep., July 2024. [Online]. Available: https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf

    2024

  3. [3]

    A face recognition method using deep learning to identify mask and unmask objects,

    S. Mishra and H. Reza, “A face recognition method using deep learning to identify mask and unmask objects,” in2022 IEEE World AI IoT Congress (AIIoT). IEEE, 2022, pp. 091–099

    2022

  4. [4]

    SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models

    P. Manakul, A. Liusie, and M. J. Gales, “Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language models,” arXiv preprint arXiv:2303.08896, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  5. [5]

    Factscore: Fine-grained atomic evaluation of factual precision in long form text generation,

    S. Min, K. Krishna, X. Lyu, M. Lewis, W.-t. Yih, P. W. Koh, M. Iyyer, L. Zettlemoyer, and H. Hajishirzi, “Factscore: Fine-grained atomic evaluation of factual precision in long form text generation,”arXiv preprint arXiv:2305.14251, 2023

    work page arXiv 2023

  6. [6]

    Ragas: Automated Evaluation of Retrieval Augmented Generation

    S. Es, J. James, L. Espinosa-Anke, and S. Schockaert, “Ragas: Au- tomated evaluation of retrieval augmented generation,”arXiv preprint arXiv:2309.15217, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  7. [7]

    Reflexion: Language Agents with Verbal Reinforcement Learning

    N. Shinn, F. Labash, A. Gopinath, and K. Narasimhan, “Reflexion: Language agents with verbal reinforcement learning,”arXiv preprint arXiv:2303.11366, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  8. [8]

    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,” inAdvances in Neural Information Processing Systems, vol. 33, 2020, pp. 9459–9474

    2020

  9. [9]

    Survey of hallucination in natural language generation,

    Z. Ji, N. Lee, R. Frieske, T. Yu, D. Su, Y . Xu, E. Ishii, Y . Bang, A. Madotto, and P. Fung, “Survey of hallucination in natural language generation,”ACM Computing Surveys, vol. 55, no. 12, pp. 1–38, 2023

    2023

  10. [10]

    Automatically correcting large language models: Survey and taxonomy,

    L. Pan, M. Saxon, R. Connor, A. Sharma, and W. Y . Wang, “Automatically correcting large language models: Survey and taxonomy,”arXiv preprint arXiv:2308.03188, 2023

    work page arXiv 2023

  11. [11]

    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

  12. [12]

    Faith and fate: Limits of transformers on compositionality,

    N. Dziri, X. Lu, M. Sclar, X. L. Li, L. Jian, B. Y . Lin, P. West, C. Bhagavatula, R. L. Bras, J. D. Hwanget al., “Faith and fate: Limits of transformers on compositionality,”Advances in Neural Information Processing Systems, vol. 36, 2023

    2023

  13. [13]

    Retrieval-Augmented Generation for Large Language Models: A Survey

    Y . Gao, Y . Xiong, X. Gao, K. Jia, J. Pan, Y . Bi, Y . Dai, J. Sun, and H. Wang, “Retrieval-augmented generation for large language models: A survey,”arXiv preprint arXiv:2312.10997, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  14. [14]

    Let's Verify Step by Step

    H. Lightmanet al., “Let’s verify step by step,”arXiv preprint arXiv:2305.20050, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  15. [15]

    Sentence-bert: Sentence embeddings using siamese bert-networks,

    N. Reimers and I. Gurevych, “Sentence-bert: Sentence embeddings using siamese bert-networks,” inProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing, 2019, pp. 3982– 3992

    2019

  16. [16]

    Deberta: Decoding-enhanced bert with disentangled attention,

    P. He, X. Liu, J. Gao, and W. Chen, “Deberta: Decoding-enhanced bert with disentangled attention,” inInternational Conference on Learning Representations, 2021

    2021

  17. [17]

    Language Models (Mostly) Know What They Know

    S. Kadavathet al., “Language models (mostly) know what they know,” arXiv preprint arXiv:2207.05221, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  18. [18]

    On calibration of modern neural networks,

    C. Guo, G. Pleiss, Y . Sun, and K. Q. Weinberger, “On calibration of modern neural networks,” inProceedings of the 34th International Conference on Machine Learning (ICML), 2017

    2017

  19. [19]

    Langchain: Building applications with llms through compos- ability,

    H. Chase, “Langchain: Building applications with llms through compos- ability,” https://github.com/hwchase17/langchain, 2023

    2023

  20. [20]

    Llamaindex: A data framework for large language models,

    J. Liu, “Llamaindex: A data framework for large language models,” https://github.com/jerryjliu/llama index, 2023

    2023

  21. [21]

    Trustworthy agentic ai pipelines: Human-in-the-loop oversight architectures for secure enterprise deployment,

    S. Mishra, “Trustworthy agentic ai pipelines: Human-in-the-loop oversight architectures for secure enterprise deployment,”ResearchGate preprint, 2026

    2026

  22. [22]

    ReAct: Synergizing reasoning and acting in language models,

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y . Cao, “ReAct: Synergizing reasoning and acting in language models,” in International Conference on Learning Representations (ICLR), 2023

    2023

  23. [23]

    Billion-scale similarity search with GPUs,

    J. Johnson, M. Douze, and H. J ´egou, “Billion-scale similarity search with GPUs,”IEEE Transactions on Big Data, 2019

    2019

  24. [24]

    Hotpotqa: A dataset for diverse, explainable multi-hop question answering,

    Z. Yang, P. Qi, S. Zhang, Y . Bengio, W. W. Cohen, R. Salakhutdinov, and C. D. Manning, “Hotpotqa: A dataset for diverse, explainable multi-hop question answering,” inProceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 2018, pp. 2369–2380

    2018

  25. [25]

    Musique: Multihop questions via single-hop question composition,

    H. Trivedi, N. Balasubramanian, T. Khot, and A. Sabharwal, “Musique: Multihop questions via single-hop question composition,” inTransactions of the Association for Computational Linguistics, vol. 10, 2022, pp. 539– 554

    2022

  26. [26]

    Constructing a multi-hop qa dataset for comprehensive evaluation of reasoning steps,

    X. Ho, A.-K. D. Nguyen, S. Sugawara, and A. Aizawa, “Constructing a multi-hop qa dataset for comprehensive evaluation of reasoning steps,” inProceedings of the 28th International Conference on Computational Linguistics, 2020, pp. 6609–6625

    2020

  27. [27]

    Ignore previous prompt: Attack techniques for language models,

    F. Perez and I. Ribeiro, “Ignore previous prompt: Attack techniques for language models,” inNeurIPS ML Safety Workshop, 2022

    2022

  28. [28]

    EVER: Mitigating hallucination in large language models through real-time verification and rectification,

    H. Kang, J. Ni, and H. Yao, “EVER: Mitigating hallucination in large language models through real-time verification and rectification,”arXiv preprint arXiv:2311.09114, 2023

    work page arXiv 2023

  29. [29]

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

    H. Trivedi, N. Balasubramanian, T. Khot, and A. Sabharwal, “Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi- step questions,” inProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL), 2023, pp. 10 014– 10 037

    2023

  30. [30]

    Agenthallu: Benchmarking automated hallucination attribution of llm-based agents,

    X. Liu, X. Yang, Z. Li, P. Li, and R. He, “Agenthallu: Benchmarking automated hallucination attribution of llm-based agents,”arXiv preprint arXiv:2601.06818, 2026

    work page arXiv 2026

  31. [31]

    Efron and R

    B. Efron and R. J. Tibshirani,An Introduction to the Bootstrap. Chapman & Hall/CRC, 1994

    1994

  32. [32]

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

    A. Asai, Z. Wu, Y . Wang, A. Salmani, and H. Hajishirzi, “Self-RAG: Learning to retrieve, generate, and critique through self-reflection,” in International Conference on Learning Representations (ICLR), 2024

    2024

  33. [33]

    Artificial intelligence risk management framework (ai rmf 1.0) (nist trustworthy and responsible ai),

    National Institute of Standards and Technology, “Artificial intelligence risk management framework (ai rmf 1.0) (nist trustworthy and responsible ai),” U.S. Department of Commerce, Tech. Rep. NIST IR 8259, January

  34. [34]

    2023 , number =

    [Online]. Available: https://doi.org/10.6028/NIST.AI.100-1

    work page doi:10.6028/nist.ai.100-1

  35. [35]

    The state of AI: How organizations are rewiring to capture value,

    A. Singla, A. Sukharevsky, L. Yee, M. Chui, and B. Hall, “The state of AI: How organizations are rewiring to capture value,” McKinsey & Company, Tech. Rep., March 2025, accessed: May 2026. [Online]. Avail- able: https://www.mckinsey.com/capabilities/quantumblack/our-insights/ the-state-of-ai-how-organizations-are-rewiring-to-capture-value

    2025

  36. [36]

    Systematization of knowledge: Security and safety in the model context protocol ecosystem,

    S. Gaire, S. Gyawali, S. Mishra, S. Niroula, D. Thakur, and U. Yadav, “Systematization of knowledge: Security and safety in the model context protocol ecosystem,”arXiv preprint arXiv:2512.08290, 2025

    work page arXiv 2025

  37. [37]

    Traq: Trustworthy retrieval augmented question answering via conformal prediction,

    S. Li, S. Park, I. Lee, and O. Bastani, “Traq: Trustworthy retrieval augmented question answering via conformal prediction,” inProceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), 2024, pp. 3799–3821

    2024

  38. [38]

    HaluEval: A large-scale hallucination evaluation benchmark for large language models,

    J. Li, X. Cheng, W. X. Zhao, J.-Y . Nie, and J.-R. Wen, “HaluEval: A large-scale hallucination evaluation benchmark for large language models,”arXiv preprint arXiv:2305.11747, 2023

    work page arXiv 2023

  39. [39]

    A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models

    S. T. I. Tonmoy, S. Zaman, V . Jain, A. Krause, T. Goswamiet al., “A comprehensive survey of hallucination mitigation techniques in large language models,”arXiv preprint arXiv:2401.01313, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  40. [40]

    Ares: An automated evaluation framework for retrieval-augmented generation systems,

    J. Saad-Falcon, O. Khattab, C. Potts, and M. Zaharia, “Ares: An automated evaluation framework for retrieval-augmented generation systems,”arXiv preprint arXiv:2311.09476, 2023

    work page arXiv 2023

  41. [41]

    March: Multi-agent reinforced self-check for llm hallucination,

    Z. Li, Y . Zhang, P. Cheng, J. Song, M. Zhou, H. Li, S. Hu, Y . Qin, E. Zhao, X. Jianget al., “March: Multi-agent reinforced self-check for llm hallucination,”arXiv preprint arXiv:2603.24579, 2026

    work page arXiv 2026

  42. [42]

    Zero-knowledge llm hallucination detection and mitigation through fine-grained cross-model consistency,

    A. Goel, D. Schwartz, and Y . Qi, “Zero-knowledge llm hallucination detection and mitigation through fine-grained cross-model consistency,” inProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, 2025, pp. 1982–1999

    2025

  43. [43]

    A survey on large language model based autonomous agents,

    L. Wang, C. Ma, X. Feng, Z. Zhang, H. Yang, J. Zhang, Z. Chen, J. Tang, X. Chen, Y . Linet al., “A survey on large language model based autonomous agents,”Frontiers of Computer Science, vol. 18, no. 6, p. 186345, 2024

    2024

  44. [44]

    Ocr hinders rag: Evaluating the cascading impact of ocr on retrieval-augmented generation,

    J. Zhang, Q. Zhang, B. Wang, L. Ouyang, Z. Wen, Y . Li, K.-H. Chow, C. He, and W. Zhang, “Ocr hinders rag: Evaluating the cascading impact of ocr on retrieval-augmented generation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 17 443–17 453

    2025

  45. [45]

    Zt-mcp: A zero-trust security architecture for mcp-connected ai agents,

    S. Mishra, “Zt-mcp: A zero-trust security architecture for mcp-connected ai agents,”ResearchGate preprint, 2026

    2026