Pith. sign in

REVIEW 4 major objections 6 minor 46 references

A training-free dual-stream system turns multimodal models into auditable industrial inspectors by pairing tool-grounded global logic with budgeted local visual search.

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 ↗

T0 review · grok-4.5

2026-07-11 23:44 UTC pith:KEPSVN4I

load-bearing objection Solid training-free dual-stream system for cold-start industrial anomaly QA; gains look real under matched budgets, but atlas quality is unmeasured and optimization-based MMAD baselines are missing. the 4 major comments →

arxiv 2607.03817 v1 pith:KEPSVN4I submitted 2026-07-04 cs.CV

Global Logic and Local Search: Dual-Stream Multimodal In-Context Learning for Verifiable Industrial Anomaly Detection

Runzhi Deng , Yundi Hu , Yiming Zhong , Zhao Wang , Xixi Liu , Hongsong Wang , Caifeng Shan , Fang Zhao This is my paper
classification cs.CV
keywords industrial anomaly detectionmultimodal in-context learningmultimodal large language modelsPart-Aware Visual-Logical Atlasdual-stream verificationtraining-free inspectionMonte Carlo tree searchverifiable diagnosis
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.

Industrial anomaly detection is a compliance audit against explicit standards, not open-ended recognition, and it must work when defective samples do not yet exist. Large multimodal models generalize in few-shot settings but often fail to ground textual standards in pixels and miss tiny defects under limited resolution. This paper claims that Global Logic and Local Search (GLLS) can fix that without training: a Part-Aware Visual-Logical Atlas organizes normal references and structured specs into the inference context; a Global & Logic Stream uses a segmentation model to extract partially checkable visual facts; and a Fine-Grained & Actions Stream uses Monte Carlo tree search to pick high-resolution evidence crops under a fixed budget. The frozen multimodal model then issues a diagnosis that stays traceable to those intermediate reports and crops. On MMAD-QA and several other industrial datasets the method reports consistent gains over matched baselines, offering a cold-start path to verifiable inspection when fine-tuning on anomalies is impossible.

Core claim

GLLS establishes that industrial anomaly detection can be performed as reference-guided multimodal in-context verification without parameter updates. Offline, a Part-Aware Visual-Logical Atlas is built from language specifications and few normal images into an executable graph of parts, normal anchors, and defect definitions. Online, a Global & Logic Stream turns atlas rules into segmentation-based checkable facts (counts, geometry, arrangement), while a Fine-Grained & Actions Stream uses MCTS under a fixed crop budget to gather local evidence. Fusing both streams lets a frozen VLM produce a final diagnosis grounded in structural logic and pixel-level crops, improving closed-set accuracy on

What carries the argument

Part-Aware Visual-Logical Atlas (PVLA) plus dual-stream verification. PVLA is a heterogeneous graph linking class, parts, normal visual anchors, and defect nodes with hierarchical, state, and contrastive edges; it is populated offline from language expansion and text-prompted segmentation of normals. Online, the Global & Logic Stream converts atlas predicates into tool-grounded facts and structural audits, and the Fine-Grained & Actions Stream treats crop selection as a budgeted MDP solved by heatmap-guided MCTS, after which the VLM verifies against the retrieved atlas context.

Load-bearing premise

The method assumes that offline expansion of product taxonomies into part and defect rules, plus filtered segmentation masks from normal images alone, yields standards accurate enough that both streams and the final model verdict can trust them when no defective examples exist.

What would settle it

On a held-out product family where atlas part/defect rules or segmentation masks systematically mismatch true component structure, GLLS accuracy under the same backbone and crop budget should fall at or below the matched base VLM on MMAD-QA closed-set discrimination and defect classification rather than show the reported gains.

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

If this is right

  • Cold-start lines can deploy multimodal inspection using only language standards and a few normal references, without defective training data or gradient updates.
  • Intermediate structural reports and evidence crops make each diagnosis auditable rather than a black-box score.
  • Budgeted MCTS local search can replace exhaustive high-resolution scanning while still lifting defect classification and localization under fixed computation.
  • The same dual-stream pattern transfers across industrial domains when atlas construction and frozen tools stay fixed.
  • Zero-shot (text-only) and one-shot (reference-guided) regimes both improve, so performance scales with available normals rather than requiring a full anomaly set.

Where Pith is reading between the lines

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

  • If atlas quality dominates, factories may need a short human review of auto-generated part and defect taxonomies before deployment rather than fully automatic expansion.
  • The same global-logic-plus-budgeted-search template could extend to other high-stakes visual compliance settings where written standards exist but defective exemplars do not.
  • Tightening the MCTS reward to VLM consistency with atlas rules, not only heatmap saliency, may cut false positives when offline standards are strict.
  • Traceable evidence chains suggest a practical path for logging automated inspection decisions for later audit without end-to-end model certification.

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 proposes GLLS, a training-free dual-stream framework for industrial anomaly detection with large multimodal models under cold-start constraints (few or no defective samples). Offline, a Part-Aware Visual-Logical Atlas (PVLA) is built by expanding MMAD seed taxonomies with GPT-5 into part/defect definitions and contrastive edges, then grounding normal references with SAM 3 text-prompted masks. Online, a Global & Logic Stream uses SAM 3 to extract partially checkable structural facts (counts, arrangement, coarse geometry), while a Fine-Grained & Actions Stream uses MCTS under a fixed crop budget, guided by a frozen anomaly heatmap prior (AdaptCLIP/ABounD), to select local evidence crops. A frozen VLM verifier fuses the structural report, crops, and hierarchical PVLA context into a closed-set answer. On MMAD-QA (MVTec/VisA) the method reports average accuracy gains over matched Qwen backbones in 0-shot and 1-shot settings; binary accuracy on MPDD/DTD/DAGM is competitive with proprietary LMMs and some specialized detectors. Ablations (Tables 3–4, 6) attribute gains to PVLA and dual-stream search under fixed backbones and budgets, and the paper emphasizes that the final diagnosis remains linked to an explicit inspection trace.

Significance. Cold-start industrial inspection with scarce defect data is a practically important setting, and a training-free, reference-guided alternative to RL/SFT pipelines is valuable if the gains hold. Strengths include matched ablations under fixed VLM backbones and evidence budgets (Tables 3–4), explicit search-policy and budget studies (Table 6), parallel latency accounting (Table 5), and cross-dataset binary evaluation beyond MMAD-QA (Table 2). The dual-stream design—tool-grounded structural checks plus budgeted local search—is a coherent systems contribution, and the paper is relatively transparent about when offline knowledge alone can hurt (e.g., Qwen2.5-7B in Table 4b). If atlas construction and intermediate fact quality were better validated, the work would be a solid systems paper for multimodal industrial AD; as written, the empirical gains are interesting but the central “verifiable grounding” claim is only partially supported.

major comments (4)
  1. [§3.2 Hybrid Atlas Construction; Eqs. (1)–(3)] §3.2 and Eqs. (1)–(3): The load-bearing claim that inspection is tool-grounded and verifiable rests on PVLA quality—GPT-5 expansion of MMAD seed taxonomies into parts/defects/contrastive edges, plus SAM 3 masks filtered by fixed confidence/IoU/stability. No experiment reports atlas fidelity (part-mask IoU vs. any part annotations, rule correctness rate, or fraction of defect definitions that are wrong or incomplete). Ablations (Tables 3–4) only remove whole modules; they do not inject controlled atlas noise or measure how often F_fact / the Logic Engine report is factually false. Without this, both streams and the closed-set verifier can inherit systematic atlas error while still producing an “evidence chain,” which weakens the verifiability claim in the abstract and contributions.
  2. [Abstract; §4.2 Main Results; Table 1] Abstract / §4.2 / Table 1: The paper repeatedly claims “consistent gains,” but Table 1 shows non-trivial per-metric drops relative to the same backbone (e.g., Loc. −1.9 for GLLS Qwen3-VL 4B 0-shot on MVTec; Desc. −2.0 and Loc. −1.4 for Qwen3-VL 8B 0-shot on VisA; several Ana./Desc. red entries). Average improvements are real under the reported protocol, but the wording should be tightened to average or majority-metric gains, and the paper should analyze when dual-stream evidence hurts localization or description rather than only highlighting green averages.
  3. [§3.3 Dual-Stream Verifiable Inference; Fig. 3] §3.3 Global & Logic Stream and final verification (Eqs. 3, 6): Intermediate structural reports are described as partially checkable, yet there is no quantitative audit of intermediate correctness (e.g., agreement of SAM-derived counts/geometry with ground truth, or rate at which a wrong Logic Engine report is overridden vs. accepted by the VLM). Figure 3 is qualitative only. For a paper whose title and framing emphasize verifiable, auditable inspection, at least one controlled study of intermediate-fact error and its effect on final closed-set accuracy is needed; otherwise “traceable” reduces to process logging rather than reliability of the evidence chain.
  4. [§4.1 Implementation Details; Table 1; Table 6] §4.1 Implementation Details and Fine-Grained stream: Local search is guided by AdaptCLIP (0-shot) or ABounD (1-shot)—specialized anomaly localizers, including prior work from the same group—while main Table 1 baselines are general-purpose VLMs without those priors. Matched GLLS ablations help, but the headline comparison still conflates dual-stream orchestration with access to a strong external AD heatmap. A control that replaces AdaptCLIP/ABounD with a generic saliency/center prior (or reports GLLS without any specialized heatmap) is needed to isolate how much of the gain is MCTS+PVLA versus the frozen specialist prior.
minor comments (6)
  1. [§5 Conclusion] Conclusion opening line is concatenated (“WepresentedGLLS,atraining-free…”); fix spacing/formatting throughout the camera-ready text.
  2. [§3.3, Eq. (4)] Eq. (4) UCB form is nonstandard in typesetting (nested fractions and the placement of λ S_search); clarify that λ=0.5 is fixed and whether S_search is normalized across states.
  3. [Table 1] Table 1 gain/loss superscripts are hard to parse in dense form; consider a separate Δ column or appendix table of absolute scores only.
  4. [§3.2; §4.1] Reproducibility: offline atlas construction depends on GPT-5 and SAM 3; state prompt templates, filtering thresholds numerically, and whether the released atlas (if any) freezes the expanded graph so results do not drift with API/model updates.
  5. [§2; §4.2] Related work cites optimization-based IAD methods extensively but Table 1 omits them on MMAD-QA; a short note on why those checkpoints cannot be evaluated under the same closed-set protocol (or a limited re-eval where possible) would reduce the appearance of selective comparison.
  6. [Fig. 1] Fig. 1 right-panel “five metrics” overview is described but quantitative values are only fully readable in Table 1; ensure figure captions state backbone and shot setting explicitly.

Circularity Check

0 steps flagged

No circularity: empirical systems paper whose accuracy claims are measured on external benchmarks, not forced by definition or self-citation.

full rationale

GLLS is a training-free dual-stream engineering framework (PVLA offline atlas + Global & Logic with SAM 3 + Fine-Grained & Actions with MCTS), not a first-principles derivation. Its load-bearing claims are closed-set QA accuracy on MMAD-QA (MVTec/VisA) and binary accuracy on MPDD/DTD/DAGM, reported against fixed ground-truth labels under matched backbones and fixed evidence budgets (Tables 1–6). Hyperparameters (N=50, Cp=2.0, Kmax=3, τsim=0.05) are chosen by ablation and do not redefine the target metrics. Offline atlas construction (GPT-5 expansion of MMAD seed taxonomies; SAM 3 text-prompted masks) supplies context for the VLM verifier; it does not fit parameters to the evaluation answers. Use of AdaptCLIP/ABounD as frozen heatmap priors for MCTS is disclosed modular tooling; even where ABounD overlaps authors, it does not justify or force the reported MMAD-QA answers. No equation equates a claimed prediction to a fitted input by construction, no uniqueness theorem is imported from the authors, and no known empirical pattern is merely renamed. The paper is self-contained against external benchmarks; circularity score is zero.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 2 invented entities

The central empirical claim rests on domain assumptions about industrial inspection (global structure + local evidence), frozen external tools (SAM 3, CLIP-style localizers, GPT-5 offline), and hand-set search/budget hyperparameters. Invented entities are methodological constructs (PVLA, dual streams), not new physical objects. No formal theorem is claimed; free parameters mainly control search cost and filtering.

free parameters (6)
  • MCTS simulation count N = 50
    Fixed at 50 per decision step after ablation; controls search thoroughness and latency.
  • MCTS exploration constant Cp = 2.0
    Set to 2.0; balances exploration vs exploitation in UCB selection.
  • Heatmap prior weight λ = 0.5
    Weights S_search in the modified UCB; set to 0.5 without broad sensitivity beyond related ablations.
  • Simulation reward gate τ_sim = 0.05
    Permissive threshold 0.05 for assigning nonzero leaf reward from heatmap score; fixed across experiments.
  • Evidence crop budget K_max = 3
    Maximum number of NMS-filtered crops passed to the verifier; chosen as 3 from budget ablation.
  • NMS IoU threshold for crop proposals = 0.1
    0.1 IoU used to enforce spatial diversity of final evidence crops.
axioms (5)
  • domain assumption Industrial anomaly detection can be usefully decomposed into tool-grounded global structural checks plus budgeted local high-resolution evidence acquisition.
    Stated as motivation via Reverse Hierarchy Theory and the dual-stream design in §1 and §3.1; load-bearing for why GLLS should beat passive VLM ICL.
  • domain assumption SAM 3 text-conditioned segmentation yields sufficiently reliable part masks and partially checkable visual facts (counts, arrangement, color order, coarse geometry) for industrial images.
    Global & Logic Stream equation (3) and §3.3 depend on Ψ_SAM producing usable F_fact.
  • ad hoc to paper GPT-5 offline expansion of MMAD seed taxonomies into sub-parts, defect modes, and contrastive edges is accurate enough after filtering to serve as executable standards.
    Hybrid Atlas Construction §3.2; no independent industrial standards audit is reported.
  • domain assumption Lightweight anomaly heatmaps from AdaptCLIP (0-shot) or ABounD (1-shot) provide a useful heuristic prior for MCTS without needing exhaustive scanning.
    Fine-Grained & Actions Stream §3.3 and Implementation Details; search quality depends on H and S_search.
  • domain assumption Closed-set MMAD-QA task accuracy under fixed answer options is an adequate primary measure of industrial inspection quality for the claimed setting.
    Setup §4.1 explicitly prefers QA accuracy over AUROC/PRAUC for the main benchmark.
invented entities (2)
  • Part-Aware Visual-Logical Atlas (PVLA) no independent evidence
    purpose: Heterogeneous graph organizing class/part/normal/defect nodes and edges to inject executable standards and references into VLM context.
    Core offline construct defined in §3.2; independent evidence is only the paper’s ablations, not an external standard atlas.
  • GLLS dual-stream verification pipeline no independent evidence
    purpose: Orchestrate Global & Logic (SAM 3 facts + auditor) and Fine-Grained & Actions (MCTS crops) into reference-guided final VLM verification.
    The proposed system as a whole; evaluated only within this paper’s experiments.

pith-pipeline@v1.1.0-grok45 · 21236 in / 3805 out tokens · 34033 ms · 2026-07-11T23:44:35.015202+00:00 · methodology

0 comments
read the original abstract

Large Multimodal Models (LMMs) show strong few-shot generalization, but industrial anomaly detection remains difficult because defects are small, input resolution is limited, and textual standards are not always grounded in visual evidence. Recent optimization-based methods improve alignment through fine-tuning, but they often require many defective samples, which are unavailable in early deployment. We present Global Logic and Local Search (GLLS), a training-free framework for reference-guided multimodal in-context verification. GLLS uses a Part-Aware Visual-Logical Atlas to organize normal references and structured specifications in the inference context. It combines a Global & Logic Stream, where SAM 3 extracts partially checkable visual facts, with a Fine-Grained & Actions Stream, where MCTS selects local evidence crops under a fixed budget. Experiments on MMAD-QA and additional anomaly detection datasets show consistent gains over matched and general-purpose baselines, while keeping the final diagnostic decision traceable to explicit visual evidence throughout the inspection trace.

Figures

Figures reproduced from arXiv: 2607.03817 by Caifeng Shan, Fang Zhao, Hongsong Wang, Runzhi Deng, Xixi Liu, Yiming Zhong, Yundi Hu, Zhao Wang.

Figure 1
Figure 1. Figure 1: Motivation and Performance Overview. The left panels visualize critical failure modes in industrial inspection, where LMMs typically suffer from Logic Hallu￾cination and Visual Bias (top), while specialized models of smaller scale struggle with Semantic Blindness and Resolution Limits (bottom). We address these limitations by incorporating a tool-grounded Global Logic stream and an active Local Search stre… view at source ↗
Figure 2
Figure 2. Figure 2: System Overview. Offline Construction: The Part-Aware Visual-Logical Atlas (PVLA) is constructed from natural language specifications and a few normal reference images. The SAM 3 segmentation model extracts visual primitives to populate the atlas with canonical visual norms and executable logic. Online Dual-Stream Inference: (1) The Global & Logic Stream executes tool-grounded checks via the SAM 3 model to… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative Visualization of the GLLS Inference Process. [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

46 extracted references · 24 linked inside Pith

  1. [1]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Aota, T., Tong, L.T.T., Okatani, T.: Zero-shot versus many-shot: Unsupervised texture anomaly detection. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 5564–5572 (2023) 16 R. Deng et al

    2023

  2. [2]

    arXiv preprint arXiv:2511.21631 (2025)

    Bai, S., Cai, Y., Chen, R., Chen, K., Chen, X., Cheng, Z., Deng, L., Ding, W., Gao, C., Ge, C., et al.: Qwen3-vl technical report. arXiv preprint arXiv:2511.21631 (2025)

    Pith/arXiv arXiv 2025

  3. [3]

    arXiv preprint arXiv:2502.13923 (2025)

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., et al.: Qwen2.5-vl technical report. arXiv preprint arXiv:2502.13923 (2025)

    Pith/arXiv arXiv 2025

  4. [4]

    Autonomous Robots42(2), 177–196 (2018)

    Bajcsy, R., Aloimonos, Y., Tsotsos, J.K.: Revisiting active perception. Autonomous Robots42(2), 177–196 (2018)

    2018

  5. [5]

    Baldassini, F.B., Shukor, M., Cord, M., Soulier, L., Piwowarski, B.: What makes multimodal in-context learning work? In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1539–1550 (2024)

    2024

  6. [6]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Bergmann, P., Fauser, M., Sattlegger, D., Steger, C.: Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9592–9600 (2019)

    2019

  7. [7]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Bigverdi, M., Luo, Z., Hsieh, C.Y., Shen, E., Chen, D., Shapiro, L.G., Krishna, R.: Perception tokens enhance visual reasoning in multimodal language models. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 3836–3845 (2025)

    2025

  8. [8]

    IEEE Transactions on Computational Intelligence and AI in games4(1), 1–43 (2012)

    Browne, C.B., Powley, E., Whitehouse, D., Lucas, S.M., Cowling, P.I., Rohlfshagen, P., Tavener, S., Perez, D., Samothrakis, S., Colton, S.: A survey of monte carlo tree search methods. IEEE Transactions on Computational Intelligence and AI in games4(1), 1–43 (2012)

    2012

  9. [9]

    arXiv preprint arXiv:2511.16719 (2025)

    Carion, N., Gustafson, L., Hu, Y.T., Debnath, S., Hu, R., Suris, D., Ryali, C., Alwala, K.V., Khedr, H., Huang, A., et al.: Sam 3: Segment anything with concepts. arXiv preprint arXiv:2511.16719 (2025)

    Pith/arXiv arXiv 2025

  10. [10]

    arXiv preprint arXiv:2504.11914 (2025)

    Chao, Y., Liu, J., Tang, J., Wu, G.: Anomalyr1: A grpo-based end-to-end mllm for industrial anomaly detection. arXiv preprint arXiv:2504.11914 (2025)

    Pith/arXiv arXiv 2025

  11. [11]

    Chen, Z., Chen, H., Imani, M., Imani, F.: Can multimodal large language mod- els be guided to improve industrial anomaly detection? In: International Design Engineering Technical Conferences and Computers and Information in Engineer- ing Conference. vol. 89213, p. V02BT02A051. American Society of Mechanical Engineers (2025)

    2025

  12. [12]

    arXiv preprint arXiv:2511.22436 (2025)

    Deng, R., Hu, Y., Zhang, X., Wang, Z., Liu, X., Dai, W.Z., Shan, C., Zhao, F.: Abound: Adversarial boundary-driven few-shot learning for multi-class anomaly detection. arXiv preprint arXiv:2511.22436 (2025)

    arXiv 2025

  13. [13]

    arXiv preprint arXiv:2505.09926 (2025)

    Gao, B.B., Zhou, Y., Yan, J., Cai, Y., Zhang, W., Wang, M., Liu, J., Liu, Y., Wang, L., Wang, C.: Adaptclip: Adapting clip for universal visual anomaly detection. arXiv preprint arXiv:2505.09926 (2025)

    Pith/arXiv arXiv 2025

  14. [14]

    arXiv preprint arXiv:2507.21619 (2025)

    Guan, W., Lan, J., Cao, J., Tan, H., Zhu, H., Wang, W.: Emit: Enhancing mllms for industrial anomaly detection via difficulty-aware grpo. arXiv preprint arXiv:2507.21619 (2025)

    Pith/arXiv arXiv 2025

  15. [15]

    arXiv preprint arXiv:2501.12948 (2025)

    Guo, D., Yang, D., Zhang, H., Song, J., Wang, P., Zhu, Q., Xu, R., Zhang, R., Ma, S., Bi, X., et al.: Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948 (2025)

    Pith/arXiv arXiv 2025

  16. [16]

    Neuron36(5), 791–804 (2002)

    Hochstein, S., Ahissar, M.: View from the top: Hierarchies and reverse hierarchies in the visual system. Neuron36(5), 791–804 (2002)

    2002

  17. [17]

    Advances in Neural Information Processing Systems37, 22124–22153 (2024) GLLS for Verifiable Industrial Anomaly Detection 17

    Huang, B., Mitra, C., Arbelle, A., Karlinsky, L., Darrell, T., Herzig, R.: Multimodal task vectors enable many-shot multimodal in-context learning. Advances in Neural Information Processing Systems37, 22124–22153 (2024) GLLS for Verifiable Industrial Anomaly Detection 17

    2024

  18. [18]

    arXiv preprint arXiv:2410.21276 (2024)

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

    Pith/arXiv arXiv 2024

  19. [19]

    arXiv preprint arXiv:2412.16720 (2024)

    Jaech, A., Kalai, A., Lerer, A., Richardson, A., El-Kishky, A., Low, A., Helyar, A., Madry, A., Beutel, A., Carney, A., et al.: Openai o1 system card. arXiv preprint arXiv:2412.16720 (2024)

    Pith/arXiv arXiv 2024

  20. [20]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Jeong, J., Zou, Y., Kim, T., Zhang, D., Ravichandran, A., Dabeer, O.: Winclip: Zero-/few-shot anomaly classification and segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 19606–19616 (2023)

    2023

  21. [21]

    In: 2021 13th International congress on ultra modern telecommunications and control systems and workshops (ICUMT)

    Jezek, S., Jonak, M., Burget, R., Dvorak, P., Skotak, M.: Deep learning-based defect detection of metal parts: evaluating current methods in complex conditions. In: 2021 13th International congress on ultra modern telecommunications and control systems and workshops (ICUMT). pp. 66–71. IEEE (2021)

    2021

  22. [22]

    arXiv preprint arXiv:2508.05503 (2025)

    Ji, D., Hu, B., Zhou, Y.: Autoiad: Manager-driven multi-agent collaboration for automated industrial anomaly detection. arXiv preprint arXiv:2508.05503 (2025)

    Pith/arXiv arXiv 2025

  23. [23]

    arXiv preprint arXiv:2410.09453 (2024)

    Jiang, X., Li, J., Deng, H., Liu, Y., Gao, B.B., Zhou, Y., Li, J., Wang, C., Zheng, F.: Mmad: A comprehensive benchmark for multimodal large language models in industrial anomaly detection. arXiv preprint arXiv:2410.09453 (2024)

    Pith/arXiv arXiv 2024

  24. [24]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Jin, E., Feng, Q., Mou, Y., Lakemeyer, G., Decker, S., Simons, O., Stegmaier, J.: Logicad: Explainable anomaly detection via vlm-based text feature extraction. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 39, pp. 4129–4137 (2025)

    2025

  25. [25]

    In: ICML

    Li, W., Fan, H., Wong, Y., Yang, Y., Kankanhalli, M.S.: Improving context under- standing in multimodal large language models via multimodal composition learning. In: ICML. vol. 3, p. 7 (2024)

    2024

  26. [26]

    arXiv preprint arXiv:2504.12749 (2025)

    Li, W., Chu, G., Chen, J., Xie, G.S., Shan, C., Zhao, F.: Lad-reasoner: Tiny multimodal models are good reasoners for logical anomaly detection. arXiv preprint arXiv:2504.12749 (2025)

    Pith/arXiv arXiv 2025

  27. [27]

    arXiv preprint arXiv:2508.09178 (2025)

    Li, Y., Cao, Y., Liu, C., Xiong, Y., Dong, X., Huang, C.: Iad-r1: Reinforcing consis- tent reasoning in industrial anomaly detection. arXiv preprint arXiv:2508.09178 (2025)

    Pith/arXiv arXiv 2025

  28. [28]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Li, Y., Yuan, S., Wang, H., Li, Q., Liu, M., Xu, C., Shi, G., Zuo, W.: Triad: Empowering lmm-based anomaly detection with expert-guided region-of-interest to- kenizer and manufacturing process. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 21917–21926 (2025)

    2025

  29. [29]

    arXiv preprint arXiv:2508.04175 (2025)

    Liao, J., Su, Y., Tu, R.C., Jin, Z., Sun, W., Li, Y., Tao, D., Xu, X., Yang, X.: Ad-fm: Multimodal llms for anomaly detection via multi-stage reasoning and fine-grained reward optimization. arXiv preprint arXiv:2508.04175 (2025)

    Pith/arXiv arXiv 2025

  30. [30]

    arXiv preprint arXiv:2404.00213 (2024)

    Mecklenburg, N., Lin, Y., Li, X., Holstein, D., Nunes, L., Malvar, S., Silva, B., Chandra, R., Aski, V., Yannam, P.K.R., et al.: Injecting new knowledge into large language models via supervised fine-tuning. arXiv preprint arXiv:2404.00213 (2024)

    Pith/arXiv arXiv 2024

  31. [31]

    arXiv preprint arXiv:2512.13671 (2025)

    Miao, J., Du, P., Liu, Y., Wang, Y., Wang, Y.: Agentiad: Tool-augmented single- agent for industrial anomaly detection. arXiv preprint arXiv:2512.13671 (2025)

    Pith/arXiv arXiv 2025

  32. [32]

    https : / / openai

    OpenAI: Introducing gpt-5. https : / / openai . com / zh - Hans - CN / index / introducing-gpt-5/(August 7 2025), accessed: 2026-01-27

    2025

  33. [33]

    Deng et al

    Song, Z., Yan, B., Liu, Y., Fang, M., Li, M., Yan, R., Chen, X.: Injecting domain- specific knowledge into large language models: A comprehensive survey (2025), https://arxiv.org/abs/2502.10708 18 R. Deng et al

    Pith/arXiv arXiv 2025

  34. [34]

    arXiv preprint arXiv:2505.15966 (2025)

    Wang, H., Su, A., Ren, W., Lin, F., Chen, W.: Pixel reasoner: Incentivizing pixel-space reasoning with curiosity-driven reinforcement learning. arXiv preprint arXiv:2505.15966 (2025)

    Pith/arXiv arXiv 2025

  35. [35]

    arXiv preprint arXiv:2508.18265 (2025)

    Wang, W., Gao, Z., Gu, L., Pu, H., Cui, L., Wei, X., Liu, Z., Jing, L., Ye, S., Shao, J., et al.: Internvl3.5: Advancing open-source multimodal models in versatility, reasoning, and efficiency. arXiv preprint arXiv:2508.18265 (2025)

    Pith/arXiv arXiv 2025

  36. [36]

    Advances in neural information processing systems35, 24824–24837 (2022)

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q.V., Zhou, D., et al.: Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems35, 24824–24837 (2022)

    2022

  37. [37]

    In: DAGM symposium in

    Wieler, M., Hahn, T.: Weakly supervised learning for industrial optical inspection. In: DAGM symposium in. vol. 6, p. 11 (2007)

    2007

  38. [38]

    arXiv preprint arXiv:2408.01800 (2024)

    Yao, Y., Yu, T., Zhang, A., Wang, C., Cui, J., Zhu, H., Cai, T., Li, H., Zhao, W., He, Z., et al.: Minicpm-v: A gpt-4v level mllm on your phone. arXiv preprint arXiv:2408.01800 (2024)

    Pith/arXiv arXiv 2024

  39. [39]

    arXiv preprint arXiv:2410.10594 (2024)

    Yu, S., Tang, C., Xu, B., Cui, J., Ran, J., Yan, Y., Liu, Z., Wang, S., Han, X., Liu, Z., et al.: Visrag: Vision-based retrieval-augmented generation on multi-modality documents. arXiv preprint arXiv:2410.10594 (2024)

    Pith/arXiv arXiv 2024

  40. [40]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Yu, T., Wang, Z., Wang, C., Huang, F., Ma, W., He, Z., Cai, T., Chen, W., Huang, Y., Zhao, R., et al.: Minicpm-v 4.5: Cooking efficient mllms via architecture, data, and training recipe. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 11704–11715 (2026)

    2026

  41. [41]

    arXiv preprint arXiv:2504.19524 (2025)

    Zeng, P., Pang, F., Wang, Z., Yang, A.: Lr-iad: Mask-free industrial anomaly detection with logical reasoning. arXiv preprint arXiv:2504.19524 (2025)

    Pith/arXiv arXiv 2025

  42. [42]

    In: 2025 IEEE International Conference on Multimedia and Expo (ICME)

    Zhang, Z., Ruan, J., Gao, X., Liu, T., Fu, Y.: Eiad: Explainable industrial anomaly detection via multi-modal large language models. In: 2025 IEEE International Conference on Multimedia and Expo (ICME). pp. 1–6. IEEE (2025)

    2025

  43. [43]

    arXiv preprint arXiv:2505.22039 (2025)

    Zhao, S., Lin, Y., Han, L., Zhao, Y., Wei, Y.: Omniad: Detect and understand industrial anomaly via multimodal reasoning. arXiv preprint arXiv:2505.22039 (2025)

    Pith/arXiv arXiv 2025

  44. [44]

    thinking with images

    Zheng, Z., Yang, M., Hong, J., Zhao, C., Xu, G., Yang, L., Shen, C., Yu, X.: Deepeyes: Incentivizing" thinking with images" via reinforcement learning. arXiv preprint arXiv:2505.14362 (2025)

    Pith/arXiv arXiv 2025

  45. [45]

    In: International Conference on Learning Representations

    Zhou, Q., Pang, G., Tian, Y., He, S., Chen, J.: Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection. In: International Conference on Learning Representations. vol. 2024, pp. 49705–49737 (2024)

    2024

  46. [46]

    In: European conference on computer vision

    Zou, Y., Jeong, J., Pemula, L., Zhang, D., Dabeer, O.: Spot-the-difference self- supervised pre-training for anomaly detection and segmentation. In: European conference on computer vision. pp. 392–408. Springer (2022)

    2022