Pith. sign in

REVIEW 3 major objections 4 minor 2 cited by

AnomalyLMM: Bridging Generative Knowledge and Discriminative Retrieval for Text-Based Person Anomaly Search

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read AnomalyLMM claims that blanking out the verbs and colors in a person-anomaly text query, having a large multimodal model fill those blanks from each candidate image's visual evidence, and semantically re-ranking the completions improves ret

desk verdict Plausible training-free LMM reranking recipe for person anomaly search, but the +0.96 R@1 gain is small and likely overfit to the test set. read the letter →

arxiv 2509.04376 v2 pith:N4UC6HOZ submitted 2025-09-04 cs.CV

Hao Ju , Hu Zhang , Zhedong Zheng This is my paper · ORCID
classification cs.CV
keywords text-basedpersonanomalysearchlargemultimodalmodelsclozecompletioncross-modalretrievalre-rankingtraining-freeadaptationdetectionPABbenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

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 sets out to show that large multimodal models can be made useful for text-based person anomaly search even though they are trained for generation, not retrieval. The proposed AnomalyLMM keeps a standard text-to-image retriever for a coarse top-N ranking, then blanks out the action verbs and color words in the query and asks an LMM to fill those blanks separately for each candidate image, using only visual evidence and 'UNKNOWN' when uncertain. A second LLM compares the completions semantically and re-ranks the candidates, and the two rankings are fused. If this works, it matters because anomaly searches are sparse and expensive to fine-tune for, so a training-free way to inject world knowledge into retrieval could improve public-safety applications. On the only public benchmark (PAB), the reported improvement over the stronger baseline is +0.96 Recall@1 and +0.46 mAP, with larger gains over a weaker retriever.

What carries the argument

The load-bearing mechanism is masked cloze completion: a prompt template with structured <VERB> and <COLOR> placeholders, filled by an LMM under strict visual-evidence instructions. The cloze does two jobs: it forces the model to localize the specific action and appearance cues named by the query instead of judging whole-image similarity, and it turns each candidate image into a short text that can be compared with the original query by an LLM in a semantically flexible way. The companion mechanism is knowledge-aware re-ranking—an LLM that counts semantically equivalent slot matches under explicit equivalence rules (same physical motion, same hue family)—whose rank is converted to a score by

What would settle it

Run the full pipeline on the PAB test set again, but feed the masked query to the LMM without showing it the candidate images (or with the image replaced by a blank caption). If Recall@1 stays close to the reported 84.73, the claimed visual grounding is not doing the work; if it drops sharply, the visual cloze is the operative mechanism.

Watch

Extended reading notes

Core claim

The central claim is that only the ground-truth image can supply the visual cues needed to fill every masked slot in an anomaly query, so the ranking should prefer images whose completions match the query semantically. AnomalyLMM operationalizes this by (1) retrieving top-N candidates with a pretrained text-to-image model, (2) replacing action verbs and color terms with <VERB>/<COLOR> placeholders using an LLM, (3) prompting an LMM to complete the placeholders for each candidate from visual evidence, with 'UNKNOWN' allowed to suppress hallucination, and (4) using an LLM to compare each completion with the original query under equivalence rules for actions and hues, then fusing the semantic r

Load-bearing premise

The true matching image must already be among the handful of candidates the ordinary text-to-image retriever returns; if it is not in that short list, the cloze and re-ranking stages cannot bring it back.

Editorial extensions

If this is right

  • Both tested coarse retrievers improve when wrapped in the pipeline (CMP +3.64 R@1, X2VLM +0.96 R@1), so the benefit is not specific to one text-to-image model.
  • Because no LMM or LLM parameters are updated, the method is deployable in sparse-data or resource-limited settings where fine-tuning a vision-language model is impractical.
  • The per-candidate cloze completions double as an explanation of the ranking: a user can see which verbs and colors matched for each image.
  • The optimal initial candidate list is short (N=3 in the experiments), which keeps the added reasoning cost bounded in practice.
  • The scheme can be layered on top of whatever fine-tuned retriever is available, so future improvements to the coarse stage can compound with it.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The gap between gains on weak and strong retrievers (+3.64 vs +0.96 R@1) suggests the cloze reranker mostly repairs coarse-stage errors; as the coarse retriever improves, the marginal gain from this exact recipe may shrink unless the masked slots are made harder or richer.
  • The UNKNOWN output offers an untapped signal: a slot-by-slot UNKNOWN rate on the true match could flag query descriptions that are too subtle for current LMMs, and might be used as a data-collection or prompt-engineering cue.
  • The same text-side compare-and-rerank design should transfer to other retrieval tasks with sparse positive samples and strong text-to-image retrievers—unusual actions, rare objects, or fine-grained attribute queries—where fine-tuning is too costly.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes AnomalyLMM, a coarse-to-fine pipeline for text-based person anomaly search on the PAB benchmark. A fine-tuned text-to-image retriever (X2VLM or CMP) produces an initial top-N ranking; an LLM then masks verbs and color terms in the query; an LMM fills those masks for each candidate image using visual evidence; an LLM compares the completions with the original query under semantic equivalence rules; and a weighted score fusion combines the initial and re-ranked lists. The central empirical claim is a +0.96 R@1 improvement (84.73 vs. 83.77) and +0.46 mAP over a fine-tuned X2VLM baseline on the PAB test set, with ablations showing both cloze completion and re-ranking contribute. The approach is described as training-free for the LMM components, though the underlying retriever is fine-tuned on PAB.

Significance. If the reported gain is robust, the paper provides a useful and interpretable recipe for leveraging off-the-shelf LMMs in fine-grained retrieval without training them. The idea of reformulating anomaly search as masked cloze completion and semantic re-ranking is novel for this task, and the qualitative analysis supports the interpretability claim. The strengths are the clear decomposition into components, the use of a public benchmark, and the ablation structure. However, the headline improvement is small (0.96 R@1 on a single 1,978-pair test split), all free hyperparameters are selected on that same test set, and no error bars or significance tests are reported; these issues currently leave the central quantitative claim vulnerable to overfitting and noise. The work is therefore significant as a proof-of-concept, but its empirical standing needs strengthening before publication.

major comments (3)
  1. [Sec. 4.2, Table 1; Tables 2(b)-(f); Eq. (1)] The headline result is a +0.96 R@1 gain over X2VLM on a single split of the PAB test set. The hyperparameters that control the method—initial candidate count N, decay base β, and fusion weights α1/α2—are all selected by ablation on this same test set (Table 2(b), 2(d), 2(e), 2(f)). Since the individual component gains in Table 2(a) are only +0.35 and +0.46 R@1, test-set selection could plausibly account for the entire reported improvement. No error bars, multiple runs, or significance tests are reported. To support the central claim, the authors should select hyperparameters on a validation split (or use nested cross-validation) and report variance or paired tests (e.g., bootstrapping over queries). This is load-bearing because the paper's main conclusion is the numerical improvement over a strong baseline.
  2. [Sec. 3.1 and Table 2(d)] The pipeline can only succeed if the ground-truth image survives the coarse top-N stage, yet the paper does not report the baseline's top-N recall for the N values considered. Table 2(d) shows N=3 is best and N=5 is worse, which suggests the re-ranker is currently the bottleneck rather than coarse recall, but this does not remove the dependence: if top-3 recall of X2VLM is well below the R@1 ceiling, the proposed re-ranking cannot recover those misses and the measured gain is contingent on a shortlist that happens to contain the target. The authors should report oracle top-N recall (for N=2..5) of the fine-tuned X2VLM and analyze failure cases where the ground-truth is outside the shortlist. Without this, the coarse-to-fine dependence is unquantified.
  3. [Sec. 4.3, Table 2(c)] The ablation over LLM/LMM choices is used to conclude that reasoning capability is important and that smaller models 'exhibit poorer performance.' However, the differences among Qwen2.5-3B, Qwen2.5-7B, Qwen3-8B, and QWQ-Max are within 0.1-0.3 R@1, which is below the likely noise level given the single-run protocol criticized above. This claim should be softened or supported by repeated runs / significance testing, especially since the final model choice (QWQ-Max) is also selected on the test set.
minor comments (4)
  1. [Table 1 and Table 2] The numbers in the 'X2VLM + Ours' rows appear duplicated ('84.73 (+0.96)84.73 (+0.96)...'), a formatting error that should be fixed.
  2. [Abstract and Sec. 4.1] The abstract and contributions call the method 'training-free,' but the pipeline requires fine-tuning the X2VLM or CMP retriever on PAB. The authors state this in Sec. 4.1, but the wording should be clarified to say that the LMM/LLM components are training-free, not the whole method.
  3. [Figure 4] The figure contains typos in the generated descriptions ('waveing', 'UNKNWON'). These should be corrected or marked as LMM outputs if intended.
  4. [Sec. 3.4, Eq. (1)] The notation S1_n and S2_n is not fully defined: it is unclear whether n indexes the same image in the initial and re-ranked lists, and how ties in the re-ranked list are assigned scores. Please clarify.

Circularity Check

1 steps flagged · score 4.0 of 10

Headline R@1/mAP are test-set-selected maxima; reported gain is not out-of-sample.

  1. fitted input called prediction [Section 4.3, Tables 2(b)-(f); Table 1]
    "...with β0.50 achieving the best performance of 84.73% Recall@1 and 90.89 mAP accuracy... Among the lengths of 2, 3, 4, and 5, re-ranking with a candidate list length of 3 achieves the best performance... varying α2 from 0.010 to 0.300 initially causes fluctuations in mAP, which peaks at 90.89 mAP when α2 = 0.075."

    Table 1 reports X2VLM+Ours as 84.73 R@1 / 90.89 mAP. These are exactly the values selected as best in the test-set ablations: β=0.50 gives 84.73/90.89, N=3 gives 84.73/90.89, and α2=0.075 gives 90.89 mAP. The hyperparameters in Eq. (1) are thus fitted to the PAB test set, and the same test set is then used to produce the headline 'prediction.' The +0.96 R@1 gain over X2VLM is therefore not an independent out-of-sample measurement; it is the maximum of a model-selection sweep on the evaluation set. This is fitted-input-called-prediction circularity: the reported performance is the value used to choose the configuration, not a prediction of that configuration's performance.

full rationale

The paper's method is a coarse-to-fine retrieval pipeline; the final ranking is a score fusion (Eq. 1) of the X2VLM initial ranking and an LLM-based re-ranking. The cloze-completion and re-ranking steps are not equivalent to the input query by construction: completions are generated from candidate images and then semantically compared to the query, so the central mechanism has independent empirical content. I find no derivation-equals-input circularity and no load-bearing self-citation chain: the PAB benchmark citation is to a public dataset, and X2VLM is an external model. However, the validation is circular in a statistical sense: the decisive hyperparameters (N, β, α1, α2) are selected by maximizing R@1/mAP on the PAB test set (Tables 2(b)-(f)), and the same test set is then used to report the headline numbers in Table 1. The reported 84.73 R@1 and 90.89 mAP are exactly the maxima found in the ablations, so the +0.96 R@1 'prediction' is a selected best configuration rather than an independent out-of-sample result. No error bars or significance tests are reported. This raises the circularity score to 4 (partial evaluation circularity) but does not make the method's derivation itself circular.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical or theoretical entities are introduced. The only new objects are prompt templates and score-fusion rules, which are not invented entities in the ledger sense.

free parameters (3)
  • Exponential decay base beta = 0.50
    Converts re-ranking position n to score beta^n; best value selected by test-set ablation, Table 2(b).
  • Score balancing weights alpha_1, alpha_2 = alpha_2 around 0.075; alpha_1 near 1 in ablations
    Final fusion weights for the score equation are tuned on PAB test set, Tables 2(e)-(f). The exact final pair is not stated clearly in the text.
  • Initial candidate count N = 3
    Number of images kept from coarse retrieval; tuned to 3 on the test set, Table 2(d).
assumptions (4)
  • domain assumption PAB test annotations are reliable ground truth for anomaly behavior.
    Used in all evaluation; no human agreement statistics are given.
  • domain assumption Fine-tuned X2VLM places the ground truth in the top-3 initial candidates for enough queries to make re-ranking meaningful.
    The method cannot recover a true match outside the candidate list; N=3 is tuned on test but top-3 recall is not reported.
  • domain assumption LMM cloze completion is grounded in visual evidence rather than textual priors despite an explicit UNKNOWN rule.
    The paper instructs the model to output UNKNOWN, but cannot verify compliance; hallucinations would corrupt re-ranking.
  • domain assumption LLM semantic equivalence judgments (e.g., sliding vs balancing) correlate with retrieval relevance.
    The comparison and re-ranking module relies entirely on this judgment (Figure 3(c)).

how reviews work

0 comments
Cite this review

Pith. "Pith review of AnomalyLMM: Bridging Generative Knowledge and Discriminative Retrieval for Text-Based Person Anomaly Search." pith.science (2026). https://pith.science/paper/N4UC6HOZ

@misc{pith2026250904376,
  author       = {Pith},
  title        = {Pith review of: AnomalyLMM: Bridging Generative Knowledge and Discriminative Retrieval for Text-Based Person Anomaly Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N4UC6HOZ}},
  note         = {Machine review of arXiv:2509.04376}
}
read the original abstract

With growing public safety demands, text-based person anomaly search has emerged as a critical task, aiming to retrieve individuals with abnormal behaviors via natural language descriptions. Unlike conventional person search, this task presents two unique challenges: (1) fine-grained cross-modal alignment between textual anomalies and visual behaviors, and (2) anomaly recognition under sparse real-world samples. While Large Multi-modal Models (LMMs) excel in multi-modal understanding, their potential for fine-grained anomaly retrieval remains underexplored, hindered by: (1) a domain gap between generative knowledge and discriminative retrieval, and (2) the absence of efficient adaptation strategies for deployment. In this work, we propose AnomalyLMM, the first framework that harnesses LMMs for text-based person anomaly search. Our key contributions are: (1) A novel coarse-to-fine pipeline integrating LMMs to bridge generative world knowledge with retrieval-centric anomaly detection; (2) A training-free adaptation cookbook featuring masked cross-modal prompting, behavioral saliency prediction, and knowledge-aware re-ranking, enabling zero-shot focus on subtle anomaly cues. As the first study to explore LMMs for this task, we conduct a rigorous evaluation on the PAB dataset, the only publicly available benchmark for text-based person anomaly search, with its curated real-world anomalies covering diverse scenarios (e.g., falling, collision, and being hit). Experiments show the effectiveness of the proposed method, surpassing the competitive baseline by +0.96% Recall@1 accuracy. Notably, our method reveals interpretable alignment between textual anomalies and visual behaviors, validated via qualitative analysis. Our code and models will be released for future research.

Figures

Figures reproduced from arXiv: 2509.04376 by the authors.

Figure 1
Figure 1. Motivation of AnomalyLMM. (Top) Direct retrieval of candidates with anomalous actions [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall pipeline of the proposed AnomalyLMM. First, given a text query about the target person with anomaly action, we apply an off-the-shelf text2image retrieval model to obtain the initial ranking list (left). Next, the verbs and colors in the text query are replaced with placeholder tokens ¡VERB¿ and ¡COLOR¿ via a pre-trained LLM, generating clozes (Cloze Generation). According to the candidate images, a pre-trai… view at source ↗
Figure 3
Figure 3. Prompt for different steps in AnomalyLMM. (a) Prompt for cloze generation. This prompt is designed for the LLM to extract verbs that describe actions from input sentences. A similar prompt is used for identifying color-related descriptions. (b) Prompt for cloze completion. This prompt guides the LMM to use visual evidence from images to fill the ¡VERB¿ and ¡COLOR¿ tokens in the masked text query with contextually ap… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Qualitative results on the PAB dataset Yang et al. (2024b). (a)-(c): successful cases; [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search

    cs.CV 2026-08 conditional novelty 6.0 of 10

    FaLCon reaches 95.41% mAP@10, 94.44% R@1, and 99.09% R@5 on PAB by combining anchor-constrained retrieval with three rerankers and an uncertainty-gated consensus.

  2. LightAIR: Lightweight Action Inversion and Riemannian Rectification for Text-based Person Anomaly Search

    cs.CV 2026-08 conditional novelty 6.0 of 10

    LightAIR combines a frozen action-word codebook, null-space appearance projection, and a Riemannian-style gradient rectification to set new state-of-the-art results on text-based person anomaly search and four TIPR be...

Reference graph

Works this paper leans on

50 extracted references · 43 canonical work pages · cited by 2 Pith papers

  1. [1]

    Gpt-4v(ision) system card. 2023. URL https://api.semanticscholar.org/CorpusID:263218031

    work page 2023

  2. [2]

    Gpt-4 technical report

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. ARXIV, 2023

    work page 2023

  3. [3]

    Flamingo: a visual language model for few-shot learning

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al. Flamingo: a visual language model for few-shot learning. NeurIPS, 35: 0 23716--23736, 2022

    work page 2022

  4. [4]

    Deep canonical correlation analysis

    Galen Andrew, Raman Arora, Jeff Bilmes, and Karen Livescu. Deep canonical correlation analysis. In ICML, pp.\ 1247--1255, 2013

    work page 2013

  5. [5]

    Rasa: Relation and sensitivity aware representation learning for text-based person search

    Yang Bai, Min Cao, Daming Gao, Ziqiang Cao, Chen Chen, Zhenfeng Fan, Liqiang Nie, and Min Zhang. Rasa: Relation and sensitivity aware representation learning for text-based person search. ARXIV, 2023

    work page 2023

  6. [6]

    Videollm-online: Online video large language model for streaming video

    Joya Chen, Zhaoyang Lv, Shiwei Wu, Kevin Qinghong Lin, Chenan Song, Difei Gao, Jia-Wei Liu, Ziteng Gao, Dongxing Mao, and Mike Zheng Shou. Videollm-online: Online video large language model for streaming video. In CVPR, pp.\ 18407--18418, 2024 a

    work page 2024

  7. [7]

    Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks

    Zhe Chen, Jiannan Wu, Wenhai Wang, Weijie Su, Guo Chen, Sen Xing, Muyan Zhong, Qinglong Zhang, Xizhou Zhu, Lewei Lu, et al. Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks. In CVPR, pp.\ 24185--24198, 2024 b

    work page 2024

  8. [8]

    Pengi: An audio language model for audio tasks

    Soham Deshmukh, Benjamin Elizalde, Rita Singh, and Huaming Wang. Pengi: An audio language model for audio tasks. NeurIPS, 36: 0 18090--18108, 2023

    work page 2023

Show all 50 references
  1. [9]

    Iterative graph attention memory network for cross-modal retrieval

    Xinfeng Dong, Huaxiang Zhang, Xiao Dong, and Xu Lu. Iterative graph attention memory network for cross-modal retrieval. Knowledge-Based Systems, 226: 0 107138, 2021

    2021

  2. [10]

    Fleet, Jamie Ryan Kiros, and Sanja Fidler

    Fartash Faghri, David J. Fleet, Jamie Ryan Kiros, and Sanja Fidler. VSE++: improving visual-semantic embeddings with hard negatives. In BMVC, pp.\ 12, 2018

    2018

  3. [11]

    Performance of 4 pre-trained sentence transformer models in the semantic query of a systematic review dataset on peri-implantitis

    Carlo Galli, Nikolaos Donos, and Elena Calciolari. Performance of 4 pre-trained sentence transformer models in the semantic query of a systematic review dataset on peri-implantitis. Information, 15 0 (2): 0 68, 2024

    2024

  4. [12]

    X-pool: Cross-modal language-video attention for text-video retrieval

    Satya Krishna Gorti, No \"e l Vouitsis, Junwei Ma, Keyvan Golestan, Maksims Volkovs, Animesh Garg, and Guangwei Yu. X-pool: Cross-modal language-video attention for text-video retrieval. In CVPR, pp.\ 5006--5015, 2022

    2022

  5. [13]

    A survey on self-supervised learning: Algorithms, applications, and future trends

    Jie Gui, Tuo Chen, Jing Zhang, Qiong Cao, Zhenan Sun, Hao Luo, and Dacheng Tao. A survey on self-supervised learning: Algorithms, applications, and future trends. IEEE TPAMI, 2024

    2024

  6. [14]

    Learning temporal regularity in video sequences

    Mahmudul Hasan, Jonghyun Choi, Jan Neumann, Amit K Roy-Chowdhury, and Larry S Davis. Learning temporal regularity in video sequences. In CVPR, pp.\ 733--742, 2016

    2016

  7. [15]

    Instruct-reid: A multi-purpose person re-identification task with instructions

    Weizhen He, Yiheng Deng, Shixiang Tang, Qihao Chen, Qingsong Xie, Yizhou Wang, Lei Bai, Feng Zhu, Rui Zhao, Wanli Ouyang, et al. Instruct-reid: A multi-purpose person re-identification task with instructions. In CVPR, pp.\ 17521--17531, 2024

    2024

  8. [16]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. ICLR, 1 0 (2): 0 3, 2022

    2022

  9. [17]

    Step-wise hierarchical alignment network for image-text matching

    Zhong Ji, Kexin Chen, and Haoran Wang. Step-wise hierarchical alignment network for image-text matching. In Zhi - Hua Zhou (ed.), IJCAI, pp.\ 765--771, 2021

    2021

  10. [18]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In ICML, pp.\ 19730--19742, 2023

    2023

  11. [19]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. NeurIPS, 36: 0 34892--34916, 2023

    2023

  12. [20]

    Rar: Retrieving and ranking augmented mllms for visual recognition

    Ziyu Liu, Zeyi Sun, Yuhang Zang, Wei Li, Pan Zhang, Xiaoyi Dong, Yuanjun Xiong, Dahua Lin, and Jiaqi Wang. Rar: Retrieving and ranking augmented mllms for visual recognition. ARXIV, 2024

    2024

  13. [21]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2019

    2019

  14. [22]

    Deep metric learning for visual understanding: An overview of recent advances

    Jiwen Lu, Junlin Hu, and Jie Zhou. Deep metric learning for visual understanding: An overview of recent advances. IEEE Signal Processing Magazine, 34 0 (6): 0 76--84, 2017

    2017

  15. [23]

    Video-chatgpt: Towards detailed video understanding via large vision and language models

    Muhammad Maaz, Hanoona Rasheed, Salman Khan, and Fahad Shahbaz Khan. Video-chatgpt: Towards detailed video understanding via large vision and language models. ARXIV, 2023

    2023

  16. [24]

    Chatreid: Open-ended interactive person retrieval via hierarchical progressive tuning for vision language models

    Ke Niu, Haiyang Yu, Mengyang Zhao, Teng Fu, Siyang Yi, Wei Lu, Bin Li, Xuelin Qian, and Xiangyang Xue. Chatreid: Open-ended interactive person retrieval via hierarchical progressive tuning for vision language models. ARXIV, 2025

    2025

  17. [25]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In ICML, pp.\ 8748--8763, 2021

    2021

  18. [26]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj \"o rn Ommer. High-resolution image synthesis with latent diffusion models. In CVPR, pp.\ 10684--10695, 2022

    2022

  19. [27]

    Backdoor cleaning without external guidance in mllm fine-tuning

    Xuankun Rong, Wenke Huang, Jian Liang, Jinhe Bi, Xun Xiao, Yiming Li, Bo Du, and Mang Ye. Backdoor cleaning without external guidance in mllm fine-tuning. ARXIV, 2025

    2025

  20. [28]

    The graph neural network model

    Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. The graph neural network model. IEEE TNNLS, 20 0 (1): 0 61--80, 2008

    2008

  21. [29]

    Semantics-enriched cross-modal alignment for complex-query video moment retrieval

    Xingyu Shen, Xiang Zhang, Xun Yang, Yibing Zhan, Long Lan, Jianfeng Dong, and Hongzhou Wu. Semantics-enriched cross-modal alignment for complex-query video moment retrieval. In ACM MM, pp.\ 4109--4118, 2023

    2023

  22. [30]

    Real-world anomaly detection in surveillance videos

    Waqas Sultani, Chen Chen, and Mubarak Shah. Real-world anomaly detection in surveillance videos. In CVPR, pp.\ 6479--6488, 2018

    2018

  23. [31]

    From data deluge to data curation: A filtering-wora paradigm for efficient text-based person search

    Jintao Sun, Hao Fei, Gangyi Ding, and Zhedong Zheng. From data deluge to data curation: A filtering-wora paradigm for efficient text-based person search. In WWW, pp.\ 2341--2351, 2025

    2025

  24. [32]

    Harnessing the power of mllms for transferable text-to-image person reid

    Wentan Tan, Changxing Ding, Jiayu Jiang, Fei Wang, Yibing Zhan, and Dapeng Tao. Harnessing the power of mllms for transferable text-to-image person reid. In CVPR, pp.\ 17127--17137, 2024

    2024

  25. [33]

    Anomaly detection in crowd scene

    Shu Wang and Zhenjiang Miao. Anomaly detection in crowd scene. In ICSPP, pp.\ 1220--1223, 2010

    2010

  26. [34]

    Cross-modal retrieval: a systematic review of methods and future directions

    Tianshi Wang, Fengling Li, Lei Zhu, Jingjing Li, Zheng Zhang, and Heng Tao Shen. Cross-modal retrieval: a systematic review of methods and future directions. Proceedings of the IEEE, 2025

    2025

  27. [35]

    Next-gpt: Any-to-any multimodal llm

    Shengqiong Wu, Hao Fei, Leigang Qu, Wei Ji, and Tat-Seng Chua. Next-gpt: Any-to-any multimodal llm. In ICML, 2024

    2024

  28. [36]

    Qwen2.5 technical report

    An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, Le Yu, Me...

    2024

  29. [37]

    Towards unified text-based person retrieval: A large-scale multi-attribute and language search benchmark

    Shuyu Yang, Yinan Zhou, Zhedong Zheng, Yaxiong Wang, Li Zhu, and Yujiao Wu. Towards unified text-based person retrieval: A large-scale multi-attribute and language search benchmark. In ACM MM, pp.\ 4492--4501, 2023

    2023

  30. [38]

    Beyond walking: A large-scale image-text benchmark for text-based person anomaly search

    Shuyu Yang, Yaxiong Wang, Li Zhu, and Zhedong Zheng. Beyond walking: A large-scale image-text benchmark for text-based person anomaly search. ARXIV, 2024 b

    2024

  31. [39]

    Minicpm-v: A gpt-4v level mllm on your phone

    Yuan Yao, Tianyu Yu, Ao Zhang, Chongyi Wang, Junbo Cui, Hongji Zhu, Tianchi Cai, Haoyu Li, Weilin Zhao, Zhihui He, et al. Minicpm-v: A gpt-4v level mllm on your phone. ARXIV, 2024

    2024

  32. [40]

    Multi-grained vision language pre-training: Aligning texts with visual concepts

    Yan Zeng, Xinsong Zhang, and Hang Li. Multi-grained vision language pre-training: Aligning texts with visual concepts. In ICML, pp.\ 25994--26009, 2022

    2022

  33. [41]

    X \^ \ 2\ 2-vlm: All-in-one pre-trained model for vision-language tasks

    Yan Zeng, Xinsong Zhang, Hang Li, Jiawei Wang, Jipeng Zhang, and Wangchunshu Zhou. X \^ \ 2\ 2-vlm: All-in-one pre-trained model for vision-language tasks. IEEE TPAMI, 46 0 (5): 0 3156--3168, 2024

    2024

  34. [42]

    Mllms know where to look: Training-free perception of small visual details with multimodal llms

    Jiarui Zhang, Mahyar Khayatkhoei, Prateek Chhikara, and Filip Ilievski. Mllms know where to look: Training-free perception of small visual details with multimodal llms. ICLR, 2025

    2025

  35. [43]

    Finetuning language models for multimodal question answering

    Xin Zhang, Wen Xie, Ziqi Dai, Jun Rao, Haokun Wen, Xuan Luo, Meishan Zhang, and Min Zhang. Finetuning language models for multimodal question answering. In ACM MM, pp.\ 9420--9424, 2023

    2023

  36. [44]

    A discriminatively learned cnn embedding for person reidentification

    Zhedong Zheng, Liang Zheng, and Yi Yang. A discriminatively learned cnn embedding for person reidentification. ACM TOMM, 14 0 (1): 0 1--20, 2017

    2017

  37. [45]

    Dual-path convolutional image-text embeddings with instance loss

    Zhedong Zheng, Liang Zheng, Michael Garrett, Yi Yang, Mingliang Xu, and Yi-Dong Shen. Dual-path convolutional image-text embeddings with instance loss. ACM TOMM, 16 0 (2): 0 1--23, 2020

    2020

  38. [46]

    Learning to prompt for vision-language models

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. Learning to prompt for vision-language models. IJCV, 130 0 (9): 0 2337--2348, 2022

    2022

  39. [47]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  40. [48]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  41. [49]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  42. [50]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.