Pith. sign in

REVIEW 3 major objections 5 minor 38 references

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

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper demonstrates that anchoring retrieval to global captions while restricting three semantic rerankers to a shared top-10 pool reaches 95.41% mAP@10, 94.44% R@1, and 99.09% R@5 on the PAB Sim2Real person anomaly search benchmark.

desk verdict Competent systems paper with official benchmark numbers; the adaptive consensus gain is real but fragile—send it to review and ask for robustness evidence. read the letter →

arxiv 2608.09474 v1 pith:BATNG54I submitted 2026-08-10 cs.CV

Hieu Dinh Trung Pham , Phuong Huu Vu Tran , Thuan Duc Mai , Son Nguyen Minh Le , Khang Le Minh , Hoang Vo , Minh-Chi Phung , Huy Minh Nhat Nguyen
show 1 more author
Cuong Tuan Nguyen
This is my paper · ORCID
classification cs.CV
keywords text-basedpersonanomalysearchSim2Realretrievalvision-languagemodelsfine-grainedimage-textmatchingcoarse-to-fineuncertainty-gatedconsensusfacetdecompositionPABbenchmark
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 text-based person anomaly search in the Sim2Real setting — models trained on about one million synthetic pairs and tested on real footage — can be solved by a coarse-to-fine cascade instead of running a single large multimodal model over the whole gallery. Its central claim is that full and concatenated captions should act as retrieval anchors while decomposed appearance, action, object, and scene facets supply bounded corrections that never introduce candidates unsupported by the anchors. Three expensive rerankers — a relevance scorer, a masked-cloze verifier, and a structured evidence verifier — then score only the shared top-10 pool, and an uncertainty-gated consensus adaptively reweights them when the fused prediction is uncertain. On the PAB benchmark the complete framework reaches 95.41% mAP@10, 94.44% R@1, and 99.09% R@5, compared with 86.44% mAP@10 for the retrieval stage alone. A reader should care because the design indicates that costly multimodal reasoning can be confined to a small candidate set without giving up recall, a pattern that matters for any retrieval task with a large gallery and a tight compute budget.

What carries the argument

The load-bearing mechanism is the uncertainty-gated consensus over a shared top-10 candidate pool, sitting on top of an anchor-constrained retriever. Three experts — a query-candidate relevance scorer from the Qwen3-VL family, a masked-cloze verifier that reconstructs masked verbs and color attributes from candidate images, and a structured Detective–Analyst–Writer verifier that synthesizes candidate-specific descriptions — each produce scores for the same ten candidates, and per-query median-IQR calibration puts them on a common scale. Static late fusion with weights $(0.3, 0.3, 0.4)$ forms a baseline belief, and a gate $g(q)$, computed from entropy, top-two margin, and pairwise Jensen-Shannon divergence of expert beliefs, interpolates between that static fusion and a credibility-weighted logarithmic opinion pool that down-weights isolated experts. The whole consensus step costs $O(|M|^2 K)$ with $|M|=3$ and $K=10$ and adds no extra multimodal forward pass.

What would settle it

Re-run the PAB official evaluation with the gate fixed to $g(q)=0$ (never adapt) and $g(q)=1$ (always adapt) using the same three rerankers. If always-adapting does not beat static fusion by roughly the reported margin, or if the validation and test histograms of $u(q)$ put the 30th and 90th percentile cut points at very different locations, then the uncertainty gate is not the source of the gain.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that a two-stage architecture — anchor-constrained soft claim-aware retrieval followed by calibrated late fusion and uncertainty-gated consensus — is enough to make Sim2Real anomaly search work at benchmark scale. Each query is decomposed into a full caption, a structured concatenation, and appearance, action, object, and scene facets; the global views act as anchors that define the admissible candidate pool, and facet branches re-rank only anchor-supported candidates through per-query percentile-based score calibration and weighted fusion. The three rerankers then score the same top-10 pool, and a credibility-weighted logarithmic opinion pool combines their beliefs while a continuous gate $g(q)$ interpolates between static fusion and the adaptive consensus using entropy, top-two margin, and pairwise Jensen-Shannon disagreement. Official-test numbers, from 86.44% mAP@10 for retrieval to 95.41% mAP@10 for the complete framework, support the paper's conclusion that preserving strong global retrieval while restricting semantic reasoning to a small pool is effective.

Load-bearing premise

The load-bearing premise is that the uncertainty-gate thresholds and fusion weights picked on the synthetic validation set transfer to real test queries; if the uncertainty distribution shifts under the Sim2Real gap, the gate could reweight the three experts incorrectly and erase the reported gains.

Editorial extensions

If this is right

  • Anchor-supported candidate generation prevents isolated facet branches from injecting unsupported candidates, so fine-grained cues can correct ranking without sacrificing recall.
  • Per-query median-IQR calibration lets heterogeneous retrievers with different score scales be fused by simple weighted sums, which is why the fused retrieval stage beats every single backbone.
  • Restricting three rerankers to a shared top-10 pool keeps semantic verification cost fixed and small, avoiding gallery-wide multimodal inference while still adding several mAP points over retrieval alone.
  • The uncertainty gate improves on static fusion (from 95.18% to 95.41% mAP@10 and from 92.47% to 94.44% R@1), so selective, query-dependent consensus is preferable to applying adaptive weights to every query.
  • Because the consensus reuses already-computed scores, any coarse-to-fine retrieval pipeline with multiple scorers on a shared pool could add it at negligible extra cost.

Reading between the lines

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

  • We infer that the uncertainty-gated consensus is expert-agnostic and could transfer to other fine-grained retrieval tasks, such as video moment retrieval or product search, whenever several rerankers score the same candidate pool.
  • We infer that the 30th and 90th percentile gate thresholds on the validation uncertainty distribution are the most fragile hyperparameters; recalibrating them on a small labeled set of real queries would directly test whether the consensus gain survives Sim2Real drift.
  • We infer that adding more experts or facets will eventually saturate; because the structured verifier gives the largest single gain in the ablation, a natural extension is to test whether a pose or temporal expert still improves consensus or mainly adds correlated agreement.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper proposes FaLCon, a coarse-to-fine framework for Sim2Real text-based person anomaly search on the PAB benchmark. The method constructs multiple textual views of a query (full caption, structured concatenation, and appearance/action/object/scene facets), fuses heterogeneous retrieval branches through anchor-constrained soft claim-aware fusion, forms a shared top-10 candidate pool, and then applies three reranking experts: a fine-tuned Qwen3 discriminative reranker, an AnomalyLMM-style cloze verifier, and an SSDC-inspired structured verifier. The expert scores are calibrated per query, combined by static late fusion, and optionally replaced by an uncertainty-gated logarithmic opinion pool that activates only on uncertain queries. The authors report official AI City 2026 Track 4 server results: soft claim-aware retrieval reaches 86.44% mAP@10, static fusion reaches 95.18% mAP@10, and the full uncertainty-gated consensus reaches 95.41% mAP@10, 94.44% R@1, and 99.09% R@5. All hyperparameters are stated to be selected on a synthetic validation split with no test-side calibration.

Significance. If the reported numbers are reliable, FaLCon is a strong empirical result on the PAB benchmark, and the paper is careful about evaluation protocol: it states that all configurations were frozen before official evaluation, reports official server scores rather than self-computed metrics, and candidly acknowledges the risk of Sim2Real calibration drift in the uncertainty gate. The component ablations are mostly monotonic and the design principle of preserving global retrieval while restricting expensive semantic reasoning to a small candidate pool is well motivated and practically attractive. The main caveat is that the final contribution, uncertainty-gated consensus, is supported by a single official test run with gate thresholds calibrated on synthetic validation, so the stability of the reported consensus gain is not yet established.

major comments (3)
  1. [Section 3.6 / Table 2] The central claim that uncertainty-gated consensus is effective rests on a single official test run whose gate thresholds δ_low and δ_high are set to the 30th and 90th percentiles of the synthetic validation u(q) distribution (Eq. 28). The paper itself warns that applying these thresholds to real queries "may remain sensitive to Sim2Real calibration drift" (Section 3.6). The consensus delta over static fusion is +0.23 mAP@10, +1.97 R@1, and +0.61 R@5, so if the real u(q) distribution is shifted upward the gate saturates at g(q)=1 and the method degenerates to a fixed log-opinion pool, while a downward shift makes consensus essentially never activate. Please provide a sensitivity analysis of the gate thresholds (e.g., varying the percentile pair or applying additive shifts to the validation u(q) distribution) and, if the evaluation protocol permits, multiple official-submission runs or a per-query breakdown showing that the adaptive-consensus gain is not a single-run artifact.
  2. [Table 2, rows 3-4] The rows '+ Qwen3 reranker' and '+ AnomalyLMM' report exactly identical R@1 (84.4287) and R@5 (97.9272). Since these are two independently implemented reranking modules, this coincidence is suspicious and, if it results from a transcription error, it undermines the component-ablation interpretation that each individual reranker improves retrieval. Please verify the official server outputs for these rows or explain the coincidence; if the row was mis-copied, correct the table and adjust the surrounding discussion.
  3. [Section 3.2 / Table 3] The 'No anchor' configuration in Table 3 is not precisely defined. The comparison between 'No anchor' and 'VM-concat anchor' conflates the effect of the anchor constraint of Eq. (3) with the effect of adding the VM-concat branch itself to the fusion, since that branch is also one of the fused retrievers. Please state explicitly which branches are active and whether the anchor constraint is applied in the 'No anchor' row, and ideally ablate the anchor constraint while keeping the same anchor branches in the fusion so that the +1.5 mAP attributed to anchoring isolates the constraint rather than the added branch.
minor comments (5)
  1. [Figure 2] The caption reads 'Overview of the our framework' and should be 'Overview of our framework'.
  2. [Section 4.2] The caption-sampling description is ambiguous: one paragraph says the VisMin CLIP is trained with Tfull/Tconcat/Tshort at 0.5/0.4/0.1, while the Qwen3 reranker paragraph says it uses mixed-caption sampling with Tfull/Tconcat at 0.5/0.5. Please state explicitly which sampling distribution applies to which model and whether the 0.5/0.4/0.1 distribution also applies to the EVA02 branch.
  3. [Equation (29)] The denominator of v_m uses d̄_ℓ, but the definition of d̄_m is given only for m; please state that d̄_ℓ is defined analogously for all ℓ ∈ M.
  4. [Section 5.3] The sentence 'The exact late-fusion fallback limits this risk' is grammatically unclear; consider 'The exact late-fusion fallback when g(q)=0 limits this risk'.
  5. [Section 6] The conclusion states that 'retrieval fusion and semantic reranking could improve performance'; 'could' should be 'can' or 'does' given that the paper reports improved official results.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central benchmark claims are validated on an independent official test set; the only same-author citation (CREDENCE) is a component, not a load-bearing proof, and the uncertainty-gate thresholds are a validation-calibrated robustness concern, not a circular fit.

full rationale

This is an empirical systems paper; its derivation chain is a cascade of trained retrievers, rerankers, and fusion rules, not an analytical derivation. Every fusion weight, temperature, and gate threshold is selected on a held-out synthetic validation split, and official test scores are reported from the AI City 2026 Track 4 server only after configurations were frozen (Section 4.1: 'All temperatures, θ_JS, δ_low, and δ_high are selected using only the held-out synthetic validation set. No test-side calibration is performed...'). Therefore the headline numbers 95.41% mAP@10, 94.44% R@1, and 99.09% R@5 are not forced by construction from validation fits; they are independent test evaluations. The uncertainty gate (Eq. 28) uses thresholds at the 30th and 90th percentiles of validation u(q); this is a hyperparameter choice, and the paper itself flags the risk ('applying these thresholds to real queries may remain sensitive to Sim2Real calibration drift', Section 3.6), which is a limitation, not circularity. The only same-author citation is CREDENCE [33], used to decompose captions into facets. The method description gives the decomposition criteria (atomicity, fidelity, entity preservation, non-redundancy), and the retrieval ablations in Table 3 show that facet-anchored retrieval improves over standalone anchors on the official test set; nothing in the paper reduces the downstream result to CREDENCE's claims by construction. The logarithmic opinion pool and uncertainty gating are standard model-combination tools cited to external sources [6,11,22]. I therefore find no circular step; the minor self-citation is not load-bearing.

Assumptions & free parameters 9 free parameters · 5 assumptions · 0 invented entities

All central hyperparameters are fitted on the synthetic validation split. No new physical or conceptual entities are introduced. The main domain assumptions are about benchmark reliability, synthetic-to-real transfer of thresholds, and the accuracy of the LLM facet decomposition.

free parameters (9)
  • Branch weights w_b in soft claim-aware retrieval = 0.35 (VisMin-Tconcat), 0.25 (VisMin-Tfull), 0.30 (Qwen3-VL-Emb.-Tfull), 0.03, 0.02, 0.006, 0.006 (facets)
    Selected on synthetic validation split; they determine the balance between anchor and corrective branches.
  • Membership term weight lambda_mem = 0.05
    Validation-selected; controls contribution of reciprocal-rank membership term.
  • Late fusion weights omega = (0.3, 0.3, 0.4) = (0.3, 0.3, 0.4)
    Grid-searched on validation for Qwen3, AnomalyLMM, SSDC experts.
  • Temperatures tau_0, tau_LF, tau_Q, tau_A, tau_S = 1.0
    Set to 1.0 for all softmax conversions.
  • Credibility sharpness theta_JS = 0.5
    Validation-selected; controls how strongly isolated experts are downweighted.
  • Uncertainty gate thresholds delta_low, delta_high = 30th/90th percentiles of validation u(q)
    Derived from validation entropy/margin/disagreement distribution.
  • Candidate pool size K = 10
    Selected from Top-5/10/20 grid on validation.
  • LoRA and training hyperparameters of Qwen3 reranker = rank=16, alpha=32, LR=1e-4, 1 epoch, 7 hard negatives
    Chosen by the authors for the fine-tuned reranker.
  • Caption sampling probabilities (0.5/0.4/0.1) = 0.5, 0.4, 0.1
    Used for Tfull/Tconcat/Tshort during VisMin-CLIP training.
assumptions (5)
  • domain assumption The PAB benchmark's official test server scores are reliable and the gallery/query pairs are correctly labeled.
    All results are reported from official eval server; the paper does not provide independent verification.
  • domain assumption The synthetic validation split is representative of the real test distribution for hyperparameter selection.
    The paper selects all thresholds and weights on a synthetic validation split and assumes transfer to real queries; it flags Sim2Real calibration drift as a sensitivity.
  • domain assumption The LLM-based facet decomposition (CREDENCE) preserves all discriminative attributes and does not introduce hallucinated claims.
    Facet quality is not directly measured; anchors mitigate but don't eliminate errors.
  • domain assumption Pretrained models (VisMin-CLIP, EVA02-CLIP, Qwen3-VL) behave as documented and are used within their intended inference procedures.
    No independent verification of these backbones is provided.
  • standard math The standard mathematical operations (softmax, KL/Jensen-Shannon divergence, median-IQR calibration) are correctly implemented.
    Formulas are standard; no proof needed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search." pith.science (2026). https://pith.science/paper/BATNG54I

@misc{pith2026260809474,
  author       = {Pith},
  title        = {Pith review of: FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BATNG54I}},
  note         = {Machine review of arXiv:2608.09474}
}
read the original abstract

Text-based person anomaly search requires retrieving real-world pedestrian images from detailed natural-language descriptions using models trained primarily on synthetic data. This Sim2Real setting is particularly challenging because visually similar candidates may differ only in subtle actions, object interactions, or appearance attributes, while applying multimodal large language models to the entire gallery is computationally expensive. We propose an anchor-constrained coarse-to-fine retrieval framework that combines global semantic matching with fine-grained verification. First, each query is represented by its original caption, a structured concatenation, and several semantic facets. Heterogeneous vision-language retrievers are then integrated through robust per-query score calibration and soft claim-aware fusion. Full and concatenated captions serve as anchors to preserve candidate recall, whereas appearance, action, and object facets provide bounded corrective evidence. The resulting candidate pool is further refined by a discriminative Qwen3 reranker and two complementary semantic verification modules based on anomaly-aware cloze completion and multi-agent evidence reasoning. Finally, an uncertainty-gated consensus module adaptively reweights the three experts on ambiguous queries. Experiments on the PAB benchmark show that the proposed soft claim-aware retrieval achieves 86.44% mAP@10, substantially outperforming individual retrieval backbones. The complete framework further improves performance to 95.41% mAP@10, 94.44% R@1, and 99.09% R@5. These results demonstrate that preserving strong global retrieval while restricting expensive semantic reasoning to a small candidate pool is effective for fine-grained Sim2Real person anomaly search. Our code will be available on Github.

Figures

Figures reproduced from arXiv: 2608.09474 by the authors.

Figure 1
Figure 1. The input description is decomposed into retrieval-oriented appearance, action, object, and scene facets. where ⊕ denotes string concatenation. The full and concatenated captions pre￾serve global semantics, while individual facets provide fine-grained evidence ( [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Overview of the our framework. Anchor-constrained multi-view retrieval constructs a shared top-K candidate pool. Three rerankers refine the candidates, and uncertainty-gated consensus combines their scores to produce the final ranking. 3.5 Calibrated Late Fusion The three rerankers score the same K = 10 candidates but produce values on different scales. We therefore normalize each expert m ∈ M := {Q, A, S} independe… view at source ↗
Figure 3
Figure 3. Uncertainty-gated consensus. Calibrated reranker scores are adaptively combined using uncertainty, disagreement, and expert credibility, reverting to standard late fusion when g(q) = 0. Because P m vm = 1, the shared prior is counted once, while candidates sup￾ported mainly by an isolated expert are suppressed. Gated output. The final belief interpolates between static fusion and adaptive consensus: \pi _i^{\mathrm … view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Anchor branches define the valid candidate pool, while corrective branches only refine the scores of anchor-supported candidates through soft claim-aware fusion [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Query-adaptive con￾sensus weighting. Low uncer￾tainty queries keep g(q) small and v ≈ ω, high uncertainty queries increase g(q) and downweight iso￾lated experts [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 27 canonical work pages

  1. [1]

    In: Advances in Neural Information Processing Systems (NeurIPS)

    Awal, R., Ahmadi, S., Zhang, L., Agrawal, A.: Vismin: Visual minimal-change understanding. In: Advances in Neural Information Processing Systems (NeurIPS). vol. 37 (2024).https://doi.org/10.52202/079017-3423

    work page doi:10.52202/079017-3423 2024

  2. [2]

    In: Proceedings of the International Joint Conference on Artificial Intelligence (IJ- CAI)

    Bai, Y., Cao, M., Gao, D., Cao, Z., Chen, C., Fan, Z., Nie, L., Zhang, M.: Rasa: Relation and sensitivity aware representation learning for text-based person search. In: Proceedings of the International Joint Conference on Artificial Intelligence (IJ- CAI). pp. 555–563 (2023)

    work page 2023

  3. [3]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    Cao,Z.,Simon,T.,Wei,S.E.,Sheikh,Y.:Realtimemulti-person2dposeestimation using part affinity fields. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 7291–7299 (2017)

    work page 2017

  4. [4]

    Neurocomputing 494, 171–181 (2022)

    Chen, Y., Zhang, G., Lu, Y., Wang, Z., Zheng, Y.: Tipcb: A simple but effective part-based convolutional baseline for text-based person search. Neurocomputing 494, 171–181 (2022)

    work page 2022

  5. [5]

    Oxford University Press (1991)

    Cooke, R.M.: Experts in Uncertainty: Opinion and Subjective Probability. Oxford University Press (1991)

    work page 1991

  6. [6]

    Journal of the American Statistical Asso- ciation69(345), 118–121 (1974)

    DeGroot, M.H.: Reaching a consensus. Journal of the American Statistical Asso- ciation69(345), 118–121 (1974)

    work page 1974

  7. [7]

    In: Proceedings of the International Conference on Machine Learning (ICML) (2024)

    Du, Y., Li, S., Torralba, A., Tenenbaum, J.B., Mordatch, I.: Improving factuality and reasoning in language models through multiagent debate. In: Proceedings of the International Conference on Machine Learning (ICML) (2024)

    work page 2024

  8. [8]

    arXiv preprint arXiv:2407.21783 (2024)

    Dubey, A., et al.: The llama 3 herd of models. arXiv preprint arXiv:2407.21783 (2024)

    arXiv 2024

Show all 38 references
  1. [9]

    arXiv preprint arXiv:2504.05047 (2025),https://arxiv.org/abs/2504.05047

    Eo, S., Moon, H., Zi, E.H., Park, C., Lim, H.: Debate only when necessary: Adaptive multiagent collaboration for efficient LLM reasoning. arXiv preprint arXiv:2504.05047 (2025),https://arxiv.org/abs/2504.05047

    2025 arXiv

  2. [10]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Fan, W., Yoon, J., Ji, B.: iMAD: Intelligent multi-agent debate for efficient and accurate LLM inference. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 40, pp. 29403–29411 (2026).https://doi.org/10.1609/aaai. v40i35.40181

    2026 doi

  3. [11]

    Statistical Science1(1), 114–135 (1986)

    Genest, C., Zidek, J.V.: Combining probability distributions: A critique and an annotated bibliography. Statistical Science1(1), 114–135 (1986)

    1986

  4. [13]

    In: International Conference on Learning Representations (2022)

    Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W.: Lora: Low-rank adaptation of large language models. In: International Conference on Learning Representations (2022)

    2022

  5. [15]

    Ilharco, G., Wortsman, M., Wightman, R., Gordon, C., Carlini, N., Taori, R., Dave, A., Shankar, V., Namkoong, H., Miller, J., Hajishirzi, H., Farhadi, A., Schmidt, L.: Openclip (2021).https://doi.org/10.5281/zenodo.7506443, software release

    2021 doi

  6. [16]

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

    Jiang, D., Ye, M.: Cross-modal implicit relation reasoning and aligning for text-to- image person retrieval. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 2787–2797 (2023) 16 H. D. T. Pham et al

    2023

  7. [17]

    In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) (2020)

    Jing, Y., Si, C., Wang, J., Wang, W., Wang, L., Tan, T.: Pose-guided multi- granularity attention network for text-based person search. In: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) (2020)

    2020

  8. [18]

    arXiv preprint arXiv:2509.04376 (2025)

    Ju, H., Zhang, H., Zheng, Z.: Anomalylmm: Bridging generative knowledge and discriminative retrieval for text-based person anomaly search. arXiv preprint arXiv:2509.04376 (2025)

    2025 arXiv

  9. [19]

    arXiv preprint arXiv:2601.04720 (2026)

    Li, M., Zhang, Y., Long, D., Chen, K., Song, S., Bai, S., Yang, Z., Xie, P., Yang, A., Liu, D., Zhou, J., Lin, J.: Qwen3-vl-embedding and qwen3-vl-reranker: A unified framework for state-of-the-art multimodal retrieval and ranking. arXiv preprint arXiv:2601.04720 (2026)

    2026 arXiv

  10. [20]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    Li, S., Xiao, T., Li, H., Zhou, B., Yue, D., Wang, X.: Person search with natural language description. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 1970–1979 (2017)

    2017

  11. [21]

    In: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)

    Liang, T., He, Z., Jiao, W., Wang, X., Wang, Y., Wang, R., Yang, Y., Shi, S., Tu, Z.: Encouraging divergent thinking in large language models through multi- agent debate. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). pp. 17889–1...

    2024

  12. [22]

    IEEE Transactions on Information Theory37(1), 145–151 (1991)

    Lin, J.: Divergence measures based on the shannon entropy. IEEE Transactions on Information Theory37(1), 145–151 (1991)

    1991

  13. [23]

    In: Com- panion Proceedings of the ACM Web Conference 2025

    Nguyen, T.H., Tran, H.L., Phan-Nguyen, H.P., Dinh, Q.V.: Hybrid, unified and iterative: A novel framework for text-based person anomaly retrieval. In: Com- panion Proceedings of the ACM Web Conference 2025. pp. 1576–1580 (2025). https://doi.org/10.1145/3701716.3717653

    2025

  14. [24]

    IEEE Transactions on Image Processing29, 5542–5556 (2020)

    Niu, K., Huang, Y., Ouyang, W., Wang, L.: Improving description-based person re-identification by multi-granularity image-text alignments. IEEE Transactions on Image Processing29, 5542–5556 (2020)

    2020

  15. [25]

    In: Proceedings of the European Conference on Computer Vision (ECCV)

    Park, J., Kim, D., Jeong, B., Kwak, S.: Plot: Text-based person search with part slot attention for corresponding part discovery. In: Proceedings of the European Conference on Computer Vision (ECCV). pp. 474–490 (2024)

    2024

  16. [26]

    In: Proceedings of the International Conference on Machine Learning (ICML)

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning trans- ferable visual models from natural language supervision. In: Proceedings of the International Conference on Machin...

    2021

  17. [27]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    Sultani, W., Chen, C., Shah, M.: Real-world anomaly detection in surveillance videos. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 6479–6488 (2018)

    2018

  18. [28]

    In: Proceedings of the ACM Web Conference (WWW)

    Sun, J., Fei, H., Ding, G., Zheng, Z.: From data deluge to data curation: A filtering- wora paradigm for efficient text-based person search. In: Proceedings of the ACM Web Conference (WWW). pp. 2341–2351 (2025)

    2025

  19. [29]

    arXiv preprint arXiv:2303.15389 (2023)

    Sun, Q., Fang, Y., Wu, L., Wang, X., Cao, Y.: Eva-clip: Improved training tech- niques for clip at scale. arXiv preprint arXiv:2303.15389 (2023)

    2023 arXiv

  20. [30]

    In: ECCV Workshops

    Tang, Z., Wang, S., Anastasiu, D.C., Chang, M.C., et al.: The 10th AI City Chal- lenge. In: ECCV Workshops. Malm"o, Sweden (2026)

    2026

  21. [31]

    arXiv preprint arXiv:2505.09388 (2025)

    Team, Q.: Qwen3 technical report. arXiv preprint arXiv:2505.09388 (2025)

    2025 arXiv

  22. [32]

    arXiv preprint arXiv:2511.21631 (2025)

    Team, Q.: Qwen3-vl technical report. arXiv preprint arXiv:2511.21631 (2025)

    2025 arXiv

  23. [33]

    arXiv preprint arXiv:2606.19819 (2026)

    Tran, P.H.V., Mai, T.D., Le, B.X.: Credence: Claim reduction for decomposition & enhanced credibility–semantic metrics and convergence analysis. arXiv preprint arXiv:2606.19819 (2026)

    2026 arXiv

  24. [34]

    In: Matu- sevych, Y., Eryiğit, G., Aletras, N

    Verma, A., Gupta, S., Gupta, D., Sircar, P., Pillai, S.: SELENE: Selective and evidence-weighted LLM debating for efficient and reliable reasoning. In: Matu- sevych, Y., Eryiğit, G., Aletras, N. (eds.) Proceedings of the 19th Conference of FaLCon 17 the European Chapter of the...

    2026 doi

  25. [35]

    In: Proceedings of the European Conference on Computer Vision (ECCV)

    Wang, Z., Fang, Z., Wang, J., Yang, Y.: Vitaa: Visual-textual attributes alignment in person search by natural language. In: Proceedings of the European Conference on Computer Vision (ECCV). pp. 402–420 (2020)

    2020

  26. [36]

    In: Findings of the Association for Computational Linguistics: ACL 2026

    Xie, Z., Luo, G., Wang, C., Cai, S., Jin, T., Zhao, Z., Tang, Y.: Bridging the pose- semantic gap: A cascade framework for text-based person anomaly search. In: Findings of the Association for Computational Linguistics: ACL 2026. pp. 4040–

    2026

  27. [37]

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

    Yang, S., Wang, Y., Zhu, L., Zheng, Z.: Beyond walking: A large-scale image- text benchmark for text-based person anomaly search. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 11720– 11730 (2025)

    2025

  28. [38]

    In: Proceedings of the International Conference on Machine Learning (ICML)

    Zeng, Y., Zhang, X., Li, H.: Multi-grained vision language pre-training: Aligning texts with visual concepts. In: Proceedings of the International Conference on Machine Learning (ICML). pp. 25994–26009 (2022)

    2022

  29. [39]

    In: Proceedings of the European Conference on Computer Vision (ECCV)

    Zhang, Y., Lu, H.: Deep cross-modal projection learning for image-text matching. In: Proceedings of the European Conference on Computer Vision (ECCV). pp. 686–701 (2018)

    2018

  30. [4049]

    Association for Computational Linguistics (2026)

    2026

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.