pith. sign in

arxiv: 2606.09641 · v1 · pith:B54UADACnew · submitted 2026-06-08 · 💻 cs.CV

MAVIS: Multi-Agent Video Retrieval via Structured Video Understanding

Jie Zhang , Qilang Ye , Hao Zhou , Haochen Liang , Fei Luo This is my paper

Pith reviewed 2026-06-27 17:25 UTC · model grok-4.3

classification 💻 cs.CV
keywords video retrievalmulti-agent frameworkstructured semantic librarylogic-aware debatecooperative reasoningattribute-level indexing
0
0 comments X

The pith

MAVIS retrieves videos by parsing them into structured semantic libraries and using multi-agent logic debate instead of full embedding scans.

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

The paper presents MAVIS as a multi-agent system that converts video retrieval from brute-force embedding search into cooperative reasoning among specialized agents. Raw videos are first parsed into a Structured Semantic Library that supports attribute-level indexing. A planner agent breaks complex queries into sub-tasks and assigns them to other agents. These agents then engage in Logic-aware Debate using a veto protocol to eliminate logical inconsistencies and surface only a small set of candidates for final verification. Experiments on MSR-VTT, MSVD, and ActivityNet show this workflow matches the accuracy of dual-encoder models while avoiding full-corpus traversal and requiring no task-specific fine-tuning.

Core claim

MAVIS rethinks video retrieval as cooperative reasoning rather than brute-force search. It parses raw videos into a Structured Semantic Library for explicit attribute-level indexing. A planner decomposes user intents into atomic sub-tasks and dispatches specialized agents. These agents then apply Logic-aware Debate with a strict veto protocol to prune logical mismatches, leaving only controversial candidates for fine-grained verification. This workflow bypasses full-library traversal and delivers competitive performance on MSR-VTT, MSVD, and ActivityNet without task-specific fine-tuning.

What carries the argument

The Logic-aware Debate mechanism, in which agents use a strict veto protocol to collaboratively eliminate logical mismatches and isolate a compact set of controversial candidates.

If this is right

  • Full-corpus embedding scans become unnecessary for video retrieval once videos are indexed in the Structured Semantic Library.
  • The veto-based debate produces an interpretable trace of why certain candidates are eliminated.
  • Competitive results hold across MSR-VTT, MSVD, and ActivityNet without any task-specific fine-tuning of the underlying models.
  • Complex user intents can be handled by decomposing them into atomic sub-tasks dispatched to independent agents.

Where Pith is reading between the lines

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

  • The same parsing-plus-debate pattern could be tested on image or text retrieval where query sparsity also creates asymmetry.
  • If the semantic library step proves stable, retrieval systems might reduce reliance on ever-larger embedding models.
  • The veto protocol itself could be reused in other multi-agent setups to improve reliability of final selections.

Load-bearing premise

Automatically parsing raw videos into a Structured Semantic Library preserves the semantic attributes needed for accurate retrieval without introducing substantial errors or omissions.

What would settle it

Retrieval accuracy on MSR-VTT or ActivityNet drops below that of standard dual-encoder baselines when the same videos are used but the semantic library parsing step is replaced by a version that introduces documented attribute omissions.

Figures

Figures reproduced from arXiv: 2606.09641 by Fei Luo, Haochen Liang, Hao Zhou, Jie Zhang, Qilang Ye.

Figure 1
Figure 1. Figure 1: Overview of MAVIS. Specialized agents tackle distinct dimensions to prune the search space through independent proposal and collaborative debate. This process identifies a minimal set of high-quality can￾didates for final fine-grained matching, ensuring both high precision and computational efficiency. This brute-force approach encounters a fundamen￾tal bottleneck: semantic asymmetry and compu￾tational red… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of MAVIS. The pipeline consists of three progressive stages: (1) Structured Video Understanding: [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Multi-agent Collaboration pipeline: (1) Query [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of MAVIS, Intern￾Video2, and Merlin. Case A: InternVideo2 suffers from retrieval biases as its global perception over-fits to background context rather than specific query intents. Case B: Merlin relies on sequential multi-round refine￾ment. MAVIS successfully retrieves the ground truth via its logic-aware veto protocol. textual queries. In cases of semantic asymmetry, InternVideo2 o… view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of Semantic Library. This figure illustrates samples from sub-libraries. Each entry maps a specific video ID to its corresponding concise semantic summary. This structured organization enables the Planner and specialized agents to perform targeted lookups rather than exhaustive scans [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
read the original abstract

The dominant paradigm in video retrieval relies on embedding-based full-corpus scanning, which suffers from inherent computational inefficiency and the semantic asymmetry between information-dense videos and sparse textual queries. To bridge this gap, we introduce \textbf{MAVIS}, a novel multi-agent framework that rethinks retrieval as cooperative reasoning rather than brute-force search. MAVIS first bridges the granularity mismatch by parsing raw videos into a \textbf{Structured Semantic Library}, enabling explicit attribute-level indexing. During retrieval, a planner decomposes complex user intents into atomic sub-tasks, dispatching specialized agents to independently nominate candidates. Crucially, MAVIS employs a \textbf{Logic-aware Debate} mechanism with a strict veto protocol, where agents collaboratively prune logical mismatches to identify a compact set of ``controversial'' candidates for fine-grained verification. This agentic workflow effectively bypasses the inefficiency of full-library traversal. Extensive experiments on MSR-VTT, MSVD, and ActivityNet demonstrate that MAVIS achieves competitive performance without task-specific fine-tuning, offering a scalable and interpretable alternative to traditional dual-encoder approaches.

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

2 major / 0 minor

Summary. The paper introduces MAVIS, a multi-agent framework for video retrieval that parses raw videos into a Structured Semantic Library for attribute-level indexing. A planner decomposes user queries into sub-tasks dispatched to specialized agents, which nominate candidates; a Logic-aware Debate mechanism with strict veto protocol then prunes logical mismatches before fine-grained verification. The central claim is that this workflow achieves competitive performance on MSR-VTT, MSVD, and ActivityNet without task-specific fine-tuning, providing a scalable and interpretable alternative to dual-encoder embedding approaches.

Significance. If substantiated, the result would offer a meaningful shift from full-corpus embedding scans to cooperative agentic reasoning, with potential gains in efficiency and interpretability for large-scale video retrieval. The no-fine-tuning aspect and explicit use of structured attributes plus debate are notable if the performance claims hold under standard protocols.

major comments (2)
  1. [Abstract] Abstract: the claim that 'extensive experiments ... demonstrate that MAVIS achieves competitive performance' supplies no quantitative results, baselines, metrics, error bars, or experimental protocol, preventing verification of the central performance claim.
  2. [Abstract and §2] Abstract and §2 (framework description): the construction of the Structured Semantic Library is described only at the conceptual level with no parser details, error analysis, or ablation on attribute fidelity (e.g., temporal relations or fine-grained actions). This is load-bearing for the claim that the library enables accurate downstream retrieval, as systematic omissions would corrupt input to the planner, agents, and debate mechanism.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments. We address each major point below and will incorporate revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'extensive experiments ... demonstrate that MAVIS achieves competitive performance' supplies no quantitative results, baselines, metrics, error bars, or experimental protocol, preventing verification of the central performance claim.

    Authors: We agree that the abstract should include quantitative support for the performance claim to allow immediate verification. In the revised manuscript we will add the key metrics (e.g., R@1, R@5, R@10 on MSR-VTT, MSVD, and ActivityNet), the main baselines, and a brief statement of the evaluation protocol while respecting abstract length constraints. revision: yes

  2. Referee: [Abstract and §2] Abstract and §2 (framework description): the construction of the Structured Semantic Library is described only at the conceptual level with no parser details, error analysis, or ablation on attribute fidelity (e.g., temporal relations or fine-grained actions). This is load-bearing for the claim that the library enables accurate downstream retrieval, as systematic omissions would corrupt input to the planner, agents, and debate mechanism.

    Authors: We acknowledge that the current description of the Structured Semantic Library remains high-level. We will expand §2 with concrete parser implementation details (model, prompting strategy, output schema), an error analysis of attribute extraction on a held-out set, and an ablation study measuring the impact of temporal-relation and action-attribute fidelity on end-to-end retrieval metrics. revision: yes

Circularity Check

0 steps flagged

No circularity: conceptual framework with no equations or self-referential derivations

full rationale

The paper describes a multi-agent video retrieval framework at the architectural level, with no equations, fitted parameters, or mathematical derivations present in the abstract or described structure. The Structured Semantic Library is introduced as an input processing step rather than defined in terms of downstream outputs. No self-citations are invoked as load-bearing uniqueness theorems, and performance claims rest on external experiments rather than internal reductions. This is a standard non-circular systems paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 2 invented entities

The central claim rests on several newly introduced conceptual components whose correctness is not independently verified in the provided text.

invented entities (2)
  • Structured Semantic Library no independent evidence
    purpose: Enable explicit attribute-level indexing of videos
    Introduced as the first processing stage to bridge granularity mismatch
  • Logic-aware Debate mechanism with strict veto protocol no independent evidence
    purpose: Prune logical mismatches among agent-nominated candidates
    Core collaboration step claimed to bypass full-library traversal

pith-pipeline@v0.9.1-grok · 5721 in / 1084 out tokens · 20030 ms · 2026-06-27T17:25:53.943768+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

63 extracted references · 9 canonical work pages · 4 internal anchors

  1. [1]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Msr-vtt: A large video description dataset for bridging video and language , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  2. [2]

    Proceedings of the IEEE international conference on computer vision , pages=

    Youtube2text: Recognizing and describing arbitrary activities using semantic hierarchies and zero-shot recognition , author=. Proceedings of the IEEE international conference on computer vision , pages=

  3. [3]

    Proceedings of the 44th International ACM SIGIR Conference on research and development in information retrieval , pages=

    Hierarchical cross-modal graph consistency learning for video-text retrieval , author=. Proceedings of the 44th International ACM SIGIR Conference on research and development in information retrieval , pages=

  4. [4]

    Proceedings of the 30th ACM International Conference on Multimedia , pages=

    Cross-lingual cross-modal retrieval with noise-robust learning , author=. Proceedings of the 30th ACM International Conference on Multimedia , pages=

  5. [5]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Unified coarse-to-fine alignment for video-text retrieval , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  6. [6]

    2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    VoP: Text-Video Co-Operative Prompt Tuning for Cross-Modal Retrieval , author=. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    2023

  7. [7]

    Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Video corpus moment retrieval with contrastive learning , author=. Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  8. [8]

    Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XXI 16 , pages=

    Tvr: A large-scale dataset for video-subtitle moment retrieval , author=. Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XXI 16 , pages=. 2020 , organization=

    2020

  9. [9]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  10. [10]

    Neurocomputing , volume=

    Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning , author=. Neurocomputing , volume=. 2022 , publisher=

    2022

  11. [11]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Wang, Jiamian and Sun, Guohao and Wang, Pichao and Liu, Dongfang and Dianat, Sohail and Rabbani, Majid and Rao, Raghuveer and Tao, Zhiqiang , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    2024

  12. [12]

    International conference on machine learning , pages=

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  13. [13]

    Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence , pages=

    Text-video retrieval with disentangled conceptualization and set-to-set alignment , author=. Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence , pages=

  14. [14]

    InternVid: A Large-scale Video-Text Dataset for Multimodal Understanding and Generation , author=

  15. [15]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Holistic features are almost sufficient for text-to-video retrieval , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  16. [16]

    arXiv preprint arXiv:2407.12508 , year=

    Merlin: Multimodal embedding refinement via llm-based iterative navigation for text-video retrieval-rerank pipeline , author=. arXiv preprint arXiv:2407.12508 , year=

    work page arXiv

  17. [17]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Text is mass: Modeling as stochastic embedding for text-video retrieval , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  18. [18]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Dgl: Dynamic global-local prompt tuning for text-video retrieval , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  19. [19]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Cap4video: What can auxiliary captions do for text-video retrieval? , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  20. [20]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Videomae v2: Scaling video masked autoencoders with dual masking , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  21. [21]

    Proceedings of the 30th ACM International Conference on Multimedia , pages=

    Partially relevant video retrieval , author=. Proceedings of the 30th ACM International Conference on Multimedia , pages=

  22. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Vita-clip: Video and text adaptive clip via multimodal prompting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  23. [23]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Timechat: A time-sensitive multimodal large language model for long video understanding , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  24. [24]

    Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding

    Video-llama: An instruction-tuned audio-visual language model for video understanding , author=. arXiv preprint arXiv:2306.02858 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  25. [25]

    Qwen2.5-VL Technical Report

    Qwen2. 5-vl technical report , author=. arXiv preprint arXiv:2502.13923 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  26. [26]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  27. [27]

    European Conference on Computer Vision , pages=

    Llama-vid: An image is worth 2 tokens in large language models , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  28. [28]

    Advances in Neural Information Processing Systems , volume=

    Self-chained image-language model for video localization and question answering , author=. Advances in Neural Information Processing Systems , volume=

  29. [29]

    International conference on machine learning , pages=

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models , author=. International conference on machine learning , pages=. 2023 , organization=

    2023

  30. [30]

    Forty-first International Conference on Machine Learning , year=

    Improving factuality and reasoning in language models through multiagent debate , author=. Forty-first International Conference on Machine Learning , year=

  31. [31]

    Advances in Neural Information Processing Systems , volume=

    Camel: Communicative agents for" mind" exploration of large language model society , author=. Advances in Neural Information Processing Systems , volume=

  32. [32]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    VarCMP: Adapting Cross-Modal Pre-Training Models for Video Anomaly Retrieval , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  33. [33]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Learning Dynamic Similarity by Bidirectional Hierarchical Sliding Semantic Probe for Efficient Text Video Retrieval , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  34. [34]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Vca: Video curious agent for long video understanding , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  35. [35]

    React: Synergizing reasoning and acting in language models , author=

  36. [36]

    Advances in Neural Information Processing Systems , volume=

    Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face , author=. Advances in Neural Information Processing Systems , volume=

  37. [37]

    Proceedings of the IEEE international conference on computer vision , pages=

    Dense-captioning events in videos , author=. Proceedings of the IEEE international conference on computer vision , pages=

  38. [38]

    European Conference on Computer Vision , year=

    InternVideo2: Scaling Foundation Models forMultimodal Video Understanding , author=. European Conference on Computer Vision , year=

  39. [39]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Muse: Mamba is efficient multi-scale learner for text-video retrieval , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  40. [40]

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

    Koala-36m: A large-scale video dataset improving consistency between fine-grained conditions and video content , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  41. [41]

    Proceedings of the 30th ACM international conference on multimedia , pages=

    X-clip: End-to-end multi-grained contrastive learning for video-text retrieval , author=. Proceedings of the 30th ACM international conference on multimedia , pages=

  42. [42]

    Advances in Neural Information Processing Systems , volume=

    Vast: A vision-audio-subtitle-text omni-modality foundation model and dataset , author=. Advances in Neural Information Processing Systems , volume=

  43. [43]

    LanguageBind: Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment

    Languagebind: Extending video-language pretraining to n-modality by language-based semantic alignment , author=. arXiv preprint arXiv:2310.01852 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  44. [44]

    2024 , howpublished =

    Introducing. 2024 , howpublished =

    2024

  45. [45]

    VideoPrism: A Foundational Visual Encoder for Video Understanding , author=

  46. [46]

    IEEE Transactions on Circuits and Systems for Video Technology , year=

    Linguistic hallucination for text-based video retrieval , author=. IEEE Transactions on Circuits and Systems for Video Technology , year=

  47. [47]

    arXiv preprint arXiv:2401.07339 , year=

    Codeagent: Enhancing code generation with tool-integrated agent systems for real-world repo-level coding challenges , author=. arXiv preprint arXiv:2401.07339 , year=

    work page arXiv

  48. [48]

    arXiv preprint arXiv:2503.10200 , year=

    Lvagent: Long video understanding by multi-round dynamical collaboration of mllm agents , author=. arXiv preprint arXiv:2503.10200 , year=

    work page arXiv

  49. [49]

    arXiv preprint arXiv:2411.13093 , year=

    Video-rag: Visually-aligned retrieval-augmented long video comprehension , author=. arXiv preprint arXiv:2411.13093 , year=

    work page arXiv

  50. [50]

    European Conference on Computer Vision , pages=

    Videoagent: A memory-augmented multimodal agent for video understanding , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  51. [51]

    Proceedings of the AAAI Conference on Artificial Intelligence , year=

    H-GAR: A Hierarchical Interaction Framework via Goal-Driven Observation-Action Refinement for Robotic Manipulation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , year=

  52. [52]

    arXiv preprint arXiv:2603.08361 , year=

    VLA: Prior-Guided Vision-Language-Action Models via World Knowledge Variation , author=. arXiv preprint arXiv:2603.08361 , year=

    work page arXiv

  53. [53]

    Proceedings of the 33nd ACM International Conference on Multimedia , year=

    EmoSym: A Symbiotic Framework for Unified Emotional Understanding and Generation via Latent Reasoning , author=. Proceedings of the 33nd ACM International Conference on Multimedia , year=

  54. [54]

    UniEmo: Unifying Emotional Understanding and Generation with Learnable Expert Queries

    UniEmo: Unifying Emotional Understanding and Generation with Learnable Expert Queries , author=. arXiv preprint arXiv:2507.23372 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  55. [55]

    2025 , eprint=

    TC-GS: Tri-plane based compression for 3D Gaussian Splatting , author=. 2025 , eprint=

    2025

  56. [56]

    European Conference on Computer Vision , pages=

    Cat: Enhancing multimodal large language model to answer questions in dynamic audio-visual scenarios , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  57. [57]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Cat+: Investigating and enhancing audio-visual understanding in large language models , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  58. [58]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    When Eyes and Ears Disagree: Can MLLMs Discern Audio-Visual Confusion? , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  59. [59]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    SUGAR: Learning Skeleton Representation with Visual-Motion Knowledge for Action Recognition , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  60. [60]

    2024 , url =

    Hai Nan and Qilang Ye and Zitong Yu and Kang An , title =. 2024 , url =

    2024

  61. [61]

    IEEE Signal Processing Letters , year=

    Pose-Promote: Progressive Visual Perception for Activities of Daily Living , author=. IEEE Signal Processing Letters , year=

  62. [62]

    2026 , eprint=

    Retrieving to Recover: Towards Incomplete Audio-Visual Question Answering via Semantic-consistent Purification , author=. 2026 , eprint=

    2026

  63. [63]

    Neural Networks , pages=

    Mgtr-miss: More ground truth retrieving based multimodal interaction and semantic supervision for video description , author=. Neural Networks , pages=. 2025 , publisher=

    2025