Pith. sign in

REVIEW 4 major objections 7 minor 56 references

LLM-powered Query Expansion for Enhancing Boundary Prediction in Language-driven Action Localization

T0 review · 4 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Adding LLM-written descriptions of an action's start and end, together with probability-based boundary supervision, improves boundary prediction across five base models and three video-moment-retrieval benchmarks.

desk verdict A well-scoped plug-in with a novel LLM-query-expansion idea, but the soft-boundary module may be silently discrete at its chosen threshold, and the evidence lacks variance controls. read the letter →

arxiv 2505.24282 v1 pith:ANKQ7T3O submitted 2025-05-30 cs.CV

Zirui Shang , Xinxiao Wu , Shuo Yang This is my paper
classification cs.CV
keywords language-drivenactionlocalizationvideomomentretrievalboundaryuncertaintyqueryexpansionlargelanguagemodelsprobabilitymodelingtemporalgrounding
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

Language-driven action localization faces a problem this paper attacks directly: the text query names the action but rarely says how the action starts or ends, so different annotators mark different frames as boundaries and the training signal is inconsistent. The paper proposes a two-part, model-agnostic recipe: prompt a large language model to write separate sentences describing the start and end of the action, then use those expanded queries to convert rigid boundary annotations into soft probability scores during training. The start/end sentences are fused into the video representation through local and global attention branches, and the probability scores are built from each frame's visual similarity to the boundary text minus its temporal distance to the annotated boundary. Integrated into five existing base models across QVHighlights, Charades-STA, and TACoS, the additions are reported to improve every evaluated metric, giving positive deltas in all 45 comparisons in the main table. The paper's claim is that this turns a coarse query and noisy boundaries into more consistent, more accurate boundary predictions without changing the base model.

What carries the argument

The machinery is a pair of plug-in modules around any existing video-query transformer. The first, LLM-powered query expansion, prompts LLaMa3-8B with a template such as "describe the beginning and ending process in one sentence of the following action", plus a constraint against inventing objects, to produce a start query $Q_s$ and an end query $Q_e$. The second, query-guided temporal modeling, lets the video features attend to $Q_s$, $Q$, and $Q_e$ in a local branch and to their concatenation in a global branch, then fuses the results into an enhanced video feature. The third, boundary probability modeling, computes a per-frame score $S^p_s(i) = \mathrm{sim}(F_{v,i}, F_s) - \mathrm{dis}(i, \tau_s)$ (and analogously for the end), takes the argmax as a pseudo-boundary, keeps only frames scoring above threshold 0.8, min-max normalizes the survivors into start/end probability curves, and sets the interior frames to probability 1. The base model is then trained with its original loss plus a cross-entropy boundary loss between its predicted frame probabilities and these soft targets.

What would settle it

Take the videos where pseudo-boundaries differ from the original annotations and ask independent annotators to mark the action start/end again. If the pseudo-boundary is not closer to the median re-annotation than the original boundary is, the core premise fails. A cheaper probe: on actions with visually subtle starts, such as "person starts laughing," check whether the pseudo-boundary lands on the true transition frame or on ordinary body motion.

Watch

Extended reading notes

Core claim

The central claim is that boundary uncertainty in language-driven action localization can be reduced without modifying the base model, by supplying the model with textual descriptions of the action start and end and by supervising it with probability scores instead of hard boundary timestamps. The paper argues that the original query under-specifies boundaries: for the same query "person eats sandwich", one annotator may mark "place hand on food plate" as the start while another marks "put food into the mouth". An LLM prompted with the action name generates canonical start/end sentences, and the boundary probability module scores every frame as $\mathrm{sim}(F_{v,i}, F_s) - \mathrm{dis}(i, \tau_s)$ for the start, and analogously for the end, takes the argmax as a pseudo-boundary, thresholds and min-max normalizes the scores into probabilities, and trains the base model with an added cross-entropy boundary loss. The authors report that all five base models improve on all metrics on all three datasets, and that the gains hold under perturbed boundary annotations. The method is presented as an off-the-shelf plug-in: only the text inputs and the training supervision change, not the architecture.

Load-bearing premise

The load-bearing premise is that the frame whose visual similarity to the LLM's start/end description, discounted by temporal distance to the human annotation, is a more reliable boundary than the human annotation itself.

Editorial extensions

If this is right

  • Any existing moment-retrieval or temporal-grounding model can adopt the recipe by changing only the text query and the training loss; no architectural re-design is needed.
  • Models trained with LLM-expanded boundary queries become more robust to annotation noise, so datasets with subjective boundary definitions become less hazardous to train on.
  • Predictions become more consistent across videos with similar queries, because the LLM supplies a canonical description of the start/end motion (e.g., "reaching out to touch the door") that anchors the boundary.
  • The choice of LLM has little effect on final performance, and even swapped start/end queries cause only a slight drop, suggesting the main benefit comes from the presence of extra boundary text and soft supervision rather than from any single generator.
  • The gains appear across all three datasets, including TACoS with dense multi-query annotations, supporting the claim that the modules are generally useful rather than dataset-specific.

Reading between the lines

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

  • A natural next experiment is to compare the pseudo-boundaries against multiple independent human boundary annotations; this would test whether the LLM's canonical start/end text is closer to the consensus boundary than a single annotator's timestamp is.
  • The same soft-supervision trick could transfer to other temporally subjective tasks such as highlight detection, phase segmentation in surgery or cooking, and weakly supervised action segmentation, wherever boundary definitions vary across annotators.
  • Because the visual features are 2-second clip-level CLIP features, the method may reach a ceiling on micro-actions with sub-second transitions; dense or motion-focused features would be a direct stress test.
  • The ablation that swaps start and end queries suggests the method may be somewhat insensitive to the exact content of the expanded text; a sharper test would be to compare LLM-generated boundary text against generic action descriptions to isolate what information actually drives the gain.
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

4 major / 7 minor

Summary. The paper proposes two plug-in modules for language-driven temporal action localization: (1) LLM-powered query expansion, which uses LLaMa3-8B to generate textual descriptions of action start and end boundaries, and (2) a query-guided temporal modeling module plus a boundary probability modeling module that converts rigid boundary annotations into soft probability labels via pseudo-boundary generation and semantic-similarity/temporal-distance scoring. The modules are inserted into five existing DETR-style base models (QD-DETR, Eatr, TaskWeave, UVCOM, CG-DETR) and evaluated on QVHighlights, Charades-STA, and TACoS, reporting consistent improvements across all 45 metric/dataset/base-model combinations in Table 1. The paper also provides component ablations, hyperparameter sensitivity studies, robustness to boundary-noise perturbations, a query-quality user study, and qualitative visualizations.

Significance. If the empirical claims hold, the method offers a simple, model-agnostic recipe for improving boundary prediction using LLM-generated text and learned soft supervision, which could be practically useful and easy to adopt. The evaluation is broad (five models, three datasets, 45 positive deltas) and the ablation structure is clean, including a useful robustness experiment to boundary annotation noise. However, the absolute gains are often small (many below 1 point), the comparisons are single-run without statistical significance or released code, and the mechanism underlying the probability-modeling module is not demonstrated to actually produce soft labels under the stated threshold. The paper is therefore valuable as an incremental contribution, but its central 'seamless and universally beneficial' claim needs stronger support.

major comments (4)
  1. [3.5, Eqs. (7)-(9), Fig. 5] The threshold τ=0.8 in Eq. (8) is applied to scores S_s(i)=sim(F_{v,i}, F_s)-dis(i,s'), where sim is a cosine similarity and dis≥0. Since any cosine similarity is at most 1, S_s(i)<1 for every frame, and for CLIP-style features typical near-boundary cosine similarities are far below 0.8. The paper does not report the fraction of frames or training samples for which S_s(i) actually exceeds τ. More tellingly, Fig. 5 shows that τ=1.0—a value at which no frame can survive the threshold because S_s(i)<1—degrades performance only modestly. This is consistent with the 'soft' probability supervision being almost always empty, so Eq. (10) degenerates to hard binary labels on [s',e']. The observed gains would then come from the pseudo-boundary shift, not from graded probability supervision. Please report survival statistics for S_s and S_e, and add a control that uses hard binary labels on the pseudo-boundary interval, to isolate the contribution of graded probabilities.
  2. [3.5, Eqs. (5)-(6), (10)] The pseudo-boundaries s' and e' are computed by independent argmaxes in Eq. (6) and are never constrained to satisfy s'≤e'. If s'>e', the piecewise definition in Eq. (10) is undefined for frames between e' and s', and the first and third branches overlap on that interval. This is a correctness gap in the training loss and could silently occur when the start query is semantically more similar to frames after the annotated start boundary or vice versa. The authors should impose an ordering constraint (e.g., clip s' to at most e') or otherwise handle this case, and report how often s'>e' occurs in practice.
  3. [4.3, Table 1] The headline claim that 'all methods integrated with our modules consistently achieve better performance on all three datasets' rests on single-run comparisons without standard deviations, significance tests, or released code for the reproduced baselines. Several deltas are small (e.g., QD-DETR R1@0.7 on QVHighlights +0.71, TaskWeave mAP on TACoS +0.71) and within typical run-to-run variation for DETR-style training. Because this is the paper's central model-agnostic claim, the authors should provide multiple seeds or bootstrap confidence intervals, and ideally release the expanded queries, code, and configuration files to enable verification of the reproduced baselines and the reported gains.
  4. [4.5, 4.6.2] The user study in Sec. 4.5 measures whether the LLM-generated start/end descriptions are relevant to the original query and consistent across similar queries, but it does not test whether those descriptions actually match the visual boundary motion in the specific video being localized. The failure case in Sec. 4.6.2 ('person starts laughing') shows that the premise fails for subtle motions. Since the mechanism of the method depends on the expanded query being visually grounded at the true boundary, the paper should acknowledge this limitation explicitly and, if possible, evaluate grounding directly (e.g., by checking whether the generated text retrieves the annotated boundary frame better than the original query).
minor comments (7)
  1. [4.3] Typo: 'moduels' should be 'modules' in the sentence 'all the methods integrated with our moduels consistently achieve better performance'.
  2. [Table 3] The row for LLaMa2-13B appears as 'LLaMa2-13B Touvron et al. (2023)' and the R1@0.7 value is printed as '38.3136.22' without a separating space; please fix the formatting and the duplicated citation.
  3. [4.4, Fig. 5] The text describing τ=1.0 says it 'filters out low-confidence frames' but does not state the stronger fact that with τ=1.0 the soft probabilities in Eq. (10) are necessarily empty because S_s(i)=sim-dis<1 for all i; this makes τ=1.0 a hard-label control and should be discussed explicitly.
  4. [3.5, Eq. (5)] The notation F_s and F_e denotes query features, but it is not specified whether these are pooled sentence-level embeddings or token-level features; clarify how the multi-token query representation is reduced to a single vector for cosine similarity.
  5. [3.4, Eq. (4)] The hyperparameters a and b are fixed to 1 in the main experiments; the sensitivity analysis in Fig. 4 covers this, but the paper should state whether the same values are used for all five base models and all three datasets, since the 'off-the-shelf' claim implies no per-model tuning.
  6. [4.1] The dataset name 'Qvhighlights' is capitalized inconsistently (QVHighlights in the abstract and elsewhere); please unify the spelling throughout.
  7. [Data Availability] Typo: 'Data Availibility' should be 'Data Availability'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the auxiliary probability supervision is anchored to human-annotated boundaries and the headline results are measured against external ground-truth metrics.

full rationale

The paper's derivation chain is not circular. The claimed contribution in Section 3.5 is a training-time label-construction rule: Eqs. (5)-(6) generate pseudo-boundaries from cosine similarity between CLIP features and LLM-expanded query features plus temporal distance to the human-annotated boundary; Eqs. (7)-(10) turn these into probability scores; Eq. (12) adds a cross-entropy loss so the base model's boundary module produces similar scores. This is auxiliary supervision, not an output that is then reported as a prediction. The headline numbers in Table 1 are R1@IoU and mAP against human ground-truth annotations on QVHighlights, Charades-STA, and TACoS, so the evaluation is external to the label-generation mechanism. L_origin (Eq. 3) is retained in Eq. (13), keeping the original ground-truth-anchored localization, saliency, and classification losses, so the final training objective is not equivalent to predicting the method's own input. There are no load-bearing self-citations: the references point to external models (CLIP, LLaMa, DETR-based base models), and the method does not invoke any prior result by the same authors as the justification for its design. The concern that the pseudo-boundary scores may degenerate under the threshold tau = 0.8 in Eq. (8) is a correctness and robustness issue about whether the soft labels are empirically active, not a circularity of the derivation chain; it does not make the reported comparisons against ground truth equal to the method's inputs. Accordingly, no specific circular step can be exhibited, and the appropriate finding is no significant circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on one fitted threshold (tau=0.8), two hand-chosen fusion weights (a=b=1), and three domain assumptions about LLM alignment, temporal proximity, and CLIP similarity validity. No invented entities are introduced: the expanded query texts are transformed inputs, not independently evidenced objects.

free parameters (3)
  • Probability threshold tau (Eq. 8) = 0.8
    Filters out low-scoring frames before min-max normalization; swept in Fig. 5 on Charades-STA and set to the reported optimum of 0.8.
  • Fusion weights a and b (Eq. 4) = a = b = 1
    Weights for global vs local branch fusion in query-guided temporal modeling; swept in Fig. 4 on Charades-STA, where equal weights performed best.
  • LLM decoding settings for query expansion = unspecified
    Temperature, sampling strategy, and generation length for LLaMa3-8B are not reported; these influence the expanded query texts and hence the supervision targets.
assumptions (5)
  • domain assumption LLM-generated start/end descriptions are semantically aligned with the actual boundary motions in each video, with hallucination controlled by the prompt restriction in Section 3.3.
    The whole pipeline's value depends on the expanded queries matching visual content; only prompt-level control is provided, with no verification against video content.
  • domain assumption Frames closer to the annotated boundary are more likely to be true boundaries, per Section 3.5: 'temporal distance reflects the intuition that frames closer to the annotated boundary are more likely to correspond to the action boundary'.
    This intuition justifies the temporal-distance penalty in Eqs. (5) and (7), anchoring pseudo-boundaries near the annotations.
  • domain assumption Cosine similarity between clip-level visual features and expanded-query text features in the frozen CLIP embedding space is a valid proxy for boundary relevance.
    Eqs. (5) and (7) rely on sim(F_v,i, F_s) at 2-second clip granularity being able to identify boundary motions; this is unverified for subtle or fast motions.
  • domain assumption The base model's boundary prediction output can be interpreted as a per-frame probability p_hat(i) across all five DETR-based architectures.
    Eq. (11) requires a frame-wise probability output, but DETR-style models output moment queries; the paper does not specify the extraction procedure for each architecture.
  • standard math Standard transformer attention, cross-entropy loss, and backpropagation work as expected.
    Background machinery of the base models and training procedure; no special mathematical assumptions beyond the method's own design.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LLM-powered Query Expansion for Enhancing Boundary Prediction in Language-driven Action Localization." pith.science (2026). https://pith.science/paper/ANKQ7T3O

@misc{pith2026250524282,
  author       = {Pith},
  title        = {Pith review of: LLM-powered Query Expansion for Enhancing Boundary Prediction in Language-driven Action Localization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ANKQ7T3O}},
  note         = {Machine review of arXiv:2505.24282}
}
read the original abstract

Language-driven action localization in videos requires not only semantic alignment between language query and video segment, but also prediction of action boundaries. However, the language query primarily describes the main content of an action and usually lacks specific details of action start and end boundaries, which increases the subjectivity of manual boundary annotation and leads to boundary uncertainty in training data. In this paper, on one hand, we propose to expand the original query by generating textual descriptions of the action start and end boundaries through LLMs, which can provide more detailed boundary cues for localization and thus reduce the impact of boundary uncertainty. On the other hand, to enhance the tolerance to boundary uncertainty during training, we propose to model probability scores of action boundaries by calculating the semantic similarities between frames and the expanded query as well as the temporal distances between frames and the annotated boundary frames. They can provide more consistent boundary supervision, thus improving the stability of training. Our method is model-agnostic and can be seamlessly and easily integrated into any existing models of language-driven action localization in an off-the-shelf manner. Experimental results on several datasets demonstrate the effectiveness of our method.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 54 canonical work pages

  1. [1]

    ArXiv preprint

    Bai J, Bai S, Chu Y, et al (2023) Qwen technical report. ArXiv preprint

    work page 2023

  2. [2]

    Computer Vision--ECCV 2020: 16th European Conference on Computer Vision (ECCV) pp 213--229

    Carion N, Massa F, Synnaeve G, et al (2020) End-to-end object detection with transformers. Computer Vision--ECCV 2020: 16th European Conference on Computer Vision (ECCV) pp 213--229

    work page 2020

  3. [3]

    In: Proc

    Chen S, Jiang Y (2019) Semantic proposal for activity localization in videos via sentence query. In: Proc. of AAAI, pp 8199--8206

    work page 2019

  4. [4]

    ArXiv preprint

    Dubey A, Jauhri A, Pandey A, et al (2024) The llama 3 herd of models. ArXiv preprint

    work page 2024

  5. [5]

    Expert Systems with Applications 269:126497

    Fang D, Xu H, Wei W, et al (2025) Multi-modal integrated proposal generation network for weakly supervised video moment retrieval. Expert Systems with Applications 269:126497

    work page 2025

  6. [6]

    In: Proc

    Feichtenhofer C, Fan H, Malik J, et al (2019) Slowfast networks for video recognition. In: Proc. of ICCV, pp 6201--6210

    work page 2019

  7. [7]

    arXiv preprint arXiv:250208544

    Flanagan K, Damen D, Wray M (2025) Moment of untruth: Dealing with negative queries in video moment retrieval. arXiv preprint arXiv:250208544

    work page 2025

  8. [8]

    In: Proc

    Gao J, Xu C (2021) Fast video moment retrieval. In: Proc. of ICCV, pp 1503--1512

    work page 2021

Show all 56 references
  1. [9]

    In: Proc

    Gao J, Sun C, Yang Z, et al (2017) TALL: temporal activity localization via language query. In: Proc. of ICCV, pp 5277--5285

    2017

  2. [10]

    In: Proc

    Gao J, Sun X, Xu M, et al (2021) Relation-aware video reading comprehension for temporal language grounding. In: Proc. of EMNLP, pp 3978--3988

    2021

  3. [11]

    In: Proc

    Ghosh S, Agarwal A, Parekh Z, et al (2019) E x CL : E xtractive C lip L ocalization U sing N atural L anguage D escriptions. In: Proc. of NAACL, pp 1984--1990

    2019

  4. [12]

    In: Proc

    Hendricks LA, Wang O, Shechtman E, et al (2017) Localizing moments in video with natural language. In: Proc. of ICCV, pp 5804--5813

    2017

  5. [13]

    In: Proc

    Huang J, Jin H, Gong S, et al (2022) Video activity localisation with uncertainties in temporal boundary. In: Proc. of ECCV, pp 724--740

    2022

  6. [14]

    Expert Systems with Applications 270:126525

    Huo S, Zhou Y, Chen K, et al (2025) Skim-and-scan transformer: A new transformer-inspired architecture for video-query based video moment retrieval. Expert Systems with Applications 270:126525

    2025

  7. [15]

    In: Proc

    Jang J, Park J, Kim J, et al (2023) Knowing where to focus: Event-aware transformer for video grounding. In: Proc. of ICCV, pp 13800--13810

    2023

  8. [16]

    In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), IEEE, pp 8586--8596

    Jung M, Jang Y, Choi S, et al (2025) Background-aware moment detection for video moment retrieval. In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), IEEE, pp 8586--8596

    2025

  9. [17]

    In: Proc

    Lei J, Berg TL, Bansal M (2021) Detecting moments and highlights in videos via natural language queries. In: Proc. of NeurIPS, pp 11846--11858

    2021

  10. [18]

    In: Proc

    Liu D, Hu W (2022) Skimming, locating, then perusing: A human-like framework for natural language video localization. In: Proc. of ACM MM, pp 4536--4545

    2022

  11. [19]

    In: Proc

    Liu D, Zhou P (2023) Jointly visual- and semantic-aware graph memory networks for temporal sentence localization in videos. In: Proc. of ICASSP, pp 1--5

    2023

  12. [20]

    In: Proc

    Liu D, Qu X, Dong J, et al (2021 a ) Adaptive proposal generation network for temporal sentence localization in videos. In: Proc. of EMNLP, pp 9292--9301

    2021

  13. [21]

    In: Proc

    Liu D, Qu X, Dong J, et al (2021 b ) Context-aware biaffine localizing network for temporal sentence grounding. In: Proc. of CVPR, pp 11235--11244

    2021

  14. [22]

    In: Proc

    Liu D, Qu X, Zhou P (2021 c ) Progressively guide to attend: An iterative alignment framework for temporal sentence grounding. In: Proc. of EMNLP, pp 9302--9311

    2021

  15. [23]

    In: Proc

    Liu D, Qu X, Di X, et al (2022 a ) Memory-guided semantic learning network for temporal sentence grounding. In: Proc. of AAAI, pp 1665--1673

    2022

  16. [24]

    In: Proc

    Liu D, Qu X, Hu W (2022 b ) Reducing the vision and language bias for temporal sentence grounding. In: Proc. of ACM MM, pp 4092--4101

    2022

  17. [25]

    IEEE Transactions on Circuits and Systems for Video Technology

    Liu J, He Z, Nie W, et al (2025) What and where: Semantic grasping and contextual scanning for moment retrieval and highlight detection. IEEE Transactions on Circuits and Systems for Video Technology

    2025

  18. [26]

    In: Proc

    Liu Z, Li J, Xie H, et al (2024) Towards balanced alignment: Modal-enhanced semantic modeling for video moment retrieval. In: Proc. of AAAI, pp 3855--3863

    2024

  19. [27]

    Knowledge-Based Systems 311:113071

    Lv Z, Su B (2025) Variational global clue inference for weakly supervised video moment retrieval. Knowledge-Based Systems 311:113071

    2025

  20. [28]

    ArXiv preprint

    Moon W, Hyun S, Lee S, et al (2023 a ) Correlation-guided query-dependency calibration in video representation learning for temporal grounding. ArXiv preprint

    2023

  21. [29]

    In: Proc

    Moon W, Hyun S, Park S, et al (2023 b ) Query - dependent video representation for moment retrieval and highlight detection. In: Proc. of CVPR, pp 23023--23033

    2023

  22. [30]

    In: Proc

    Mun J, Cho M, Han B (2020) Local-global video-text interactions for temporal grounding. In: Proc. of CVPR, pp 10807--10816

    2020

  23. [31]

    In: Proc

    Nishimura T, Nakada S, Munakata H, et al (2024) Lighthouse: A user-friendly library for reproducible video moment retrieval and highlight detection. In: Proc. of EMNLP

    2024

  24. [32]

    In: Proc

    Otani M, Nakashima Y, Rahtu E, et al (2020) Uncovering hidden challenges in query-based video moment retrieval. In: Proc. of BMVC

    2020

  25. [33]

    IEEE Transactions on Neural Networks and Learning Systems

    Pan W, Zhao Z, Huang W, et al (2022) Video moment retrieval with noisy labels. IEEE Transactions on Neural Networks and Learning Systems

    2022

  26. [34]

    In: Proc

    Radford A, Kim JW, Hallacy C, et al (2021) Learning transferable visual models from natural language supervision. In: Proc. of ICML, pp 8748--8763

    2021

  27. [35]

    TACL pp 25--36

    Regneri M, Rohrbach M, Wetzel D, et al (2013) Grounding action descriptions in videos. TACL pp 25--36

    2013

  28. [36]

    In: Proc

    Rohrbach M, Regneri M, Andriluka M, et al (2012) Script data for attribute-based recognition of composite activities. In: Proc. of ECCV, pp 144--157

    2012

  29. [37]

    In: Proc

    Sigurdsson GA, Varol G, Wang X, et al (2016) Hollywood in homes: Crowdsourcing data collection for activity understanding. In: Proc. of ECCV, pp 510--526

    2016

  30. [38]

    Mobile Networks and Applications pp 1--34

    Sun G, Liang L, Li T, et al (2021) Video question answering: a survey of models and datasets. Mobile Networks and Applications pp 1--34

    2021

  31. [39]

    In: Proc

    Sun X, Wang X, Gao J, et al (2022) You need to read again: Multi-granularity perception network for moment retrieval in videos. In: Proc. of SIGIR, pp 1022--1032

    2022

  32. [40]

    IEEE Transactions on Multimedia pp 1338--1349

    Tang H, Zhu J, Liu M, et al (2021) Frame-wise cross-modal matching for video moment retrieval. IEEE Transactions on Multimedia pp 1338--1349

    2021

  33. [41]

    ArXiv preprint

    Touvron H, Lavril T, Izacard G, et al (2023) Llama: Open and efficient foundation language models. ArXiv preprint

    2023

  34. [42]

    IEEE Transactions on Multimedia pp 1221--1232

    Wang G, Xu X, Shen F, et al (2022) Cross-modal dynamic networks for video moment retrieval with text query. IEEE Transactions on Multimedia pp 1221--1232

    2022

  35. [43]

    In: Proc

    Wu Z, Gao J, Huang S, et al (2021) Diving into the relations: Leveraging semantic and visual structures for video moment retrieval. In: Proc. of ICME, pp 1--6

    2021

  36. [44]

    In: Proc

    Xiao Y, Luo Z, Liu Y, et al (2024) Bridging the gap: A unified video comprehension framework for moment retrieval and highlight detection. In: Proc. of CVPR, pp 18709--18719

    2024

  37. [45]

    In: Proc

    Xu H, Das A, Saenko K (2017) R-C3D: region convolutional 3d network for temporal activity detection. In: Proc. of ICCV, pp 5794--5803

    2017

  38. [46]

    In: Proc

    Xu H, He K, Plummer BA, et al (2019) Multilevel language and vision integration for text-to-clip retrieval. In: Proc. of AAAI, pp 9062--9069

    2019

  39. [47]

    In: Proc

    Yang J, Wei P, Li H, et al (2024) Task-driven exploration: Decoupling and inter-task feedback for joint moment retrieval and highlight detection. In: Proc. of CVPR, pp 18308--18318

    2024

  40. [48]

    In: Proc

    Yuan Y, Ma L, Wang J, et al (2019) Semantic conditioned dynamic modulation for temporal sentence grounding in videos. In: Proc. of NeurIPS, pp 534--544

    2019

  41. [49]

    In: Proc

    Zeng Y, Cao D, Wei X, et al (2021) Multi-modal relational graph for cross-modal video moment retrieval. In: Proc. of CVPR, pp 2215--2224

    2021

  42. [50]

    In: Proc

    Zhang H, Sun A, Jing W, et al (2020 a ) Span-based localizing network for natural language video localization. In: Proc. of ACL, pp 6543--6554

    2020

  43. [51]

    IEEE Transactions on Pattern Analysis and Machine Intelligence pp 4252--4266

    Zhang H, Sun A, Jing W, et al (2022 a ) Natural language video localization: A revisit in span-based question answering framework. IEEE Transactions on Pattern Analysis and Machine Intelligence pp 4252--4266

    2022

  44. [52]

    In: Proc

    Zhang S, Peng H, Fu J, et al (2020 b ) Learning 2d temporal adjacent networks for moment localization with natural language. In: Proc. of AAAI, pp 12870--12877

    2020

  45. [53]

    Preprintsorg

    Zhang X, Li Y, Han Y, et al (2022 b ) Ai video editing: A survey. Preprintsorg

    2022

  46. [54]

    In: Proc

    Zhang Y, Chen X, Jia J, et al (2023) Text-visual prompting for efficient 2d temporal video grounding. In: Proc. of CVPR, pp 14794--14804

    2023

  47. [55]

    sn-basic.bst

    FUNCTION identify.basic.version "sn-basic.bst" " [2024/07/19 v1.1 bibliography style]" * top ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key keywords month note number organization pages publisher school series ...

    2024

  48. [56]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.