Pith. sign in

REVIEW 4 major objections 6 minor 42 references

NBA_Streaming: A Large-Scale Benchmark for Fine-Grained Basketball Commentary Generation in Continuous Streams

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read NBA_Streaming turns live basketball commentary into a measurable task, and a causal two-stage model beats existing baselines on it.

desk verdict A large, genuinely new streaming basketball benchmark whose PBP-derived boundaries and machine-generated references need scrutiny before the reported numbers can be trusted. read the letter →

arxiv 2608.09200 v2 pith:7KT5WND4 submitted 2026-08-10 cs.CV

Lifang Wu , Yuyang Wu , Yangdong Gao , Fengyu Liu , Ya Jing , Liang Wang This is my paper · ORCID
classification cs.CV
keywords basketballcommentarygenerationonlinetemporalactionlocalizationstreamingvideounderstandingfine-grainedrecognitionvideo-languagemodelingplayeridentityeventchainsbenchmarkdataset
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

Live basketball commentary is normally studied on pre-cut clips, where the model already knows what happened. This paper claims that the real task is streaming: the system must decide, from frames seen so far, when an event has fully finished and then describe it before the next event begins. To make that task measurable, it introduces NBA_Streaming, a benchmark of 307.5 hours of NBA broadcasts with roughly 35,000 temporally aligned events, each annotated with boundaries, player identities, 71 fine-grained action categories, event chains, and 50.4K natural commentary sentences. On this benchmark, the paper's causal two-stage framework—a completion-first event localizer followed by ball-centric semantic grounding—improves over existing streaming and clip-based baselines in boundary accuracy, factual grounding, and commentary quality, while leaving a clear gap on complex multi-step events. A sympathetic reader would take the central claim to be that continuous-stream commentary is a distinct, learnable task and that this benchmark plus framework is a credible first step.

What carries the argument

Two mechanisms carry the argument. The first is completion-first causal event localization: a learnable end query watches the stream and, when its end probability first crosses a threshold, fixes the current frame as the event end; then, because the start is still unknown, a completeness head scores candidate intervals of 2 to 20 seconds that all end at that frame and picks the one with the highest completeness against a temporal-IoU target. This is what lets the system respond in real time without waiting for future frames or delayed boundary merging. The second is ball-centric semantic grounding: a frozen ball detector supplies position, confidence, velocity, and acceleration, around which the model gathers visual features at three scales (the ball, its interaction context, and the ball handler), and a conditional semantic pyramid turns those cues into event-type, player-identity, and fine-grained-action prompts for the multimodal language model that writes the commentary. Identity recognition is weakly supervised: the commentary gives player names and roles but not timestamps, so the model learns to weight ball-anchored observations by role-specific queries. These two mechanisms together produce the reported gains.

What would settle it

A decisive check would be to take a fresh random sample of games outside the original 20-game manual verification, have basketball-literate annotators mark event start/end boundaries and player identities directly from video, and compare those human-consensus windows with the OCR-aligned play-by-play windows. If agreement on timestamp inclusion or complete-event coverage drops substantially below the reported 97% and 96%, or if the proposed system's identity accuracy on those games falls far below the reported 28.5%, then the alignment assumption—and with it the benchmark's ground truth—fails.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that online basketball commentary can be decomposed into two causally constrained decisions—when an event has become complete and what to say about it—and that both decisions can be supervised from one dataset built by aligning official play-by-play logs with broadcast video. The paper claims NBA_Streaming is the first benchmark to provide this coupling at scale: full-game continuous streams rather than isolated clips, event boundaries rather than clip labels, and semantic annotations deep enough to include player identities, fine-grained actions, and causally linked event chains. The proposed framework is offered as evidence that the benchmark is tractable: committing to an event at the first valid end trigger and then choosing the most complete interval from history reduces start and end boundary errors to 2.470 s and 2.094 s, while ball-centric semantic grounding lifts identity, event-type, and fine-grained action accuracy in generated commentary far above the baselines. In the paper's framing, the remaining shortfall on complex chains such as putback sequences does not weaken the benchmark but instead shows that the task is not yet saturated.

Load-bearing premise

The load-bearing premise is that the official play-by-play timestamps, aligned to broadcast time by OCR and expanded into windows by hand-written event-type rules, mark the true start and end of each event; because only 20 of the 152 games were manually checked (with 96–99% pass rates), a systematic alignment failure in the unchecked games would corrupt every boundary, identity, action, and commentary metric that follows.

Editorial extensions

If this is right

  • If the benchmark holds, future commentary systems can be compared under one continuous-stream protocol that jointly scores when to speak (localization and response recall/precision) and what to say (language quality and factual accuracy), with future frames forbidden.
  • Because the localizer commits at the first valid end trigger, its predictions arrive at event completion rather than after offline merging, so the 230.65 FPS throughput claim means the full pipeline can serve multiple live streams in real time.
  • The weakly supervised identity module shows that player names can be learned from commentary text plus ball-anchored video, without per-player region annotations, lowering the cost of identity-aware sports captioning.
  • The clear failure on multi-stage event chains (miss, offensive rebound, putback) defines a concrete research target: maintaining role and identity consistency across causally linked sub-events.
  • Substantial headroom remains—identity accuracy 28.5%, fine-grained action accuracy 47.1%—so NBA_Streaming is not saturated and can support incremental progress in streaming video understanding.

Reading between the lines

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

  • My inference: the completion-first principle is not basketball-specific; any domain with discrete events that become recognizable only after completion—surgical steps, assembly-line operations, live news sequences—could use an end trigger plus completeness-scored candidates for just-in-time narration, and the paper's protocol gives a template for such benchmarks.
  • My inference: because identity supervision comes from commentary alone, the same method could be applied to historical archives where no player tracking exists, effectively mining identity labels from the spoken or written commentary of past games.
  • My inference: the causal setting invites a new diagnostic the paper does not report—stratifying commentary accuracy by local event density; if accuracy degrades as inter-event intervals shrink, then response latency, not perception, is the binding constraint, and engineering should focus on faster generators rather than better grounding.
  • My inference: a transfer test on soccer, where official event logs and broadcast clocks also exist, would clarify whether the two-stage design, rather than basketball-specific priors, is responsible for the reported gains.
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

4 major / 6 minor

Summary. The paper introduces NBA_Streaming, a benchmark built from 152 NBA games (307.5 hours, ~35K events) with play-by-play-derived event boundaries, identity/action annotations, and natural-language commentary generated by Llama3-8B from PBP records. It also proposes a causal two-stage model: Stage I localizes completed events in an online fashion using an end-trigger and completeness scoring over end-anchored candidates, and Stage II feeds ball-centric event, identity, and action prompts to Qwen3-VL for commentary generation. Experiments compare against streaming and localization baselines, reporting lower boundary errors, higher response accuracy, higher commentary metrics, and higher throughput.

Significance. If valid, the dataset would be a useful resource for streaming sports video understanding, and the framework demonstrates a sensible way to couple online localization with commentary generation. The paper provides a detailed taxonomy, a clear evaluation protocol, and an architecture that is described in enough detail to reproduce. The significance is conditional on the temporal ground truth and the reference commentary being trustworthy; the current validation does not yet establish that.

major comments (4)
  1. [III.A, Fig. 2, V.B] The ground-truth event windows are constructed by expanding official PBP timestamps with event-type-specific boundary rules, and the manual validation on 20 games (pass rates 97%, 99%, 96%) checks timestamp inclusion, commentary consistency, and complete-event coverage, but it never measures boundary precision. There is no inter-annotator agreement on start/end times and no comparison against visually observable event boundaries. Because the matching criterion in Section V.B requires both predicted boundaries to be within ±2 s of ground truth, and because Stage II commentary metrics are computed only on events satisfying that criterion, any systematic offset or over-wide/under-wide window in the PBP-derived boundaries propagates directly into all localization and commentary numbers. This is the central validity assumption of the benchmark and needs either additional validation or an explicit analysis of expected boundary noise.
  2. [III.A] The reference commentary is generated by Llama3-8B from PBP abbreviations using event-type templates, rather than being human-written broadcast commentary. Consequently, the reported CIDEr/BLEU-4 scores and Identity/Event/Action Accuracy measure agreement with an LLM-expanded version of the same PBP record that also defines the event type and boundaries; this is a partially circular evaluation. Please clarify what skill the benchmark is actually testing and provide either a human-written reference set or a human evaluation of the reference commentaries themselves, otherwise the 'natural-language commentary' claim is overstated.
  3. [V.B, V.C, Table III] Commentary quality is evaluated only on events whose start and end errors are both within 2 s, and Table III further restricts to events commonly matched by all methods. The text in Section V.C states that frozen Qwen3-VL matches only 52 events, so if this baseline is included in the definition of 'all methods' the common set cannot exceed 52 events; in any case the table does not report the number of common events. Results on such a small and non-random subset are unlikely to be representative of full-test-set commentary quality, and the comparison may be biased by which events each method happened to localize correctly. Please report the common-event count and also give commentary metrics on the full event set or on a fixed, independent set.
  4. [V.A.2, V.C, Table II, V.C.3] All quantitative results are single-run point estimates with no error bars, and the human evaluation uses only two raters without reporting inter-annotator agreement. Because the headline claims are large margins (e.g., CIDEr 248.355 versus 73.772 for the strongest baseline in Table II), it is important to know whether these differences are stable across random seeds and rating noise. Please report multiple runs with variance measures and rater-agreement statistics.
minor comments (6)
  1. [III.A] The boundary rules are described only as 'event-type-specific'; a table or supplementary material listing the rules per event type would aid reproducibility and allow reviewers to judge the expected window sizes.
  2. [V.B] The definitions of Event Response and Correct Response are given by analogy to recall and precision, but no formal formula is provided; please add the exact set definitions.
  3. [Table II] Please state whether the reported FPS values are aggregate or per-stream, since the text says aggregate throughput is computed across three GPUs at a combined input rate of 18 FPS.
  4. [Throughout] The dataset name is spelled inconsistently as NBA_Streaming and NBA Streaming; please standardize the spelling.
  5. [Fig. 5] The identifiers such as '0022500100 / Q4 6:57.0' are not explained; please add a caption note describing the game, quarter, and time format.
  6. [III.C] The ethics statement does not address the copyright status of the broadcast videos under CC BY-NC; a clarification of data provenance and licensing would be useful.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity found: the benchmark construction and model evaluation are empirical, and no reported prediction reduces to a fitted input or self-citation.

full rationale

NBA_Streaming is an empirical benchmark and method-evaluation paper; its claims are supported by dataset construction, manual quality checks, and comparative experiments rather than by a derivation chain. The only potentially self-referential element is that the natural-language commentary used as ground truth is generated by Llama3-8B from the same official play-by-play records that seed event boundaries (Section III.A: 'we use Llama3-8B [32] to expand the PBP descriptions into natural commentary according to templates designed for different event types'). This is data construction, not circular reasoning: the evaluated models receive video (not PBP text) and are scored against that fixed reference; there is no equation in which a predicted quantity is defined in terms of the target quantity. Stage I's completeness supervision uses temporal IoU with the annotated windows as the training target, and Stage II's metrics are computed on matched events; these are standard supervised evaluations, not fitted parameters renamed as predictions. The paper does not invoke a self-authored uniqueness theorem, nor does it smuggle an ansatz via self-citation; cited prior work [9,14] is used for baseline comparison. Accordingly, no load-bearing step reduces by construction to its own inputs, and the correct circularity score is 0.

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

The central evaluation rests on several unverified assumptions about dataset construction and model components. The most notable are the accuracy of OCR time alignment, the acceptability of Llama3-8B-generated commentary as ground truth, and the reliability of the frozen ball detector. Because code and data are not released, these assumptions cannot currently be checked independently.

free parameters (3)
  • Event-ending trigger threshold (tau) = 0.5
    Stage I commits to an event the first time the predicted ending probability exceeds this threshold. The value is chosen by hand and directly controls response timing and localization.
  • End-anchored candidate event durations = 2 to 20 seconds at 0.5 second intervals, 37 candidates
    The completeness scorer selects among these fixed durations. Events shorter than 2 seconds or longer than 20 seconds cannot be recovered by the design.
  • Input sampling rate and context window = 6 FPS, 60 current frames plus 72 historical frames (about 22 seconds)
    This design choice balances latency and context. A longer or shorter window would change which plays can be localized and described.
assumptions (5)
  • domain assumption PBP records are accurate and OCR-based alignment to broadcast time is correct.
    Section III-A: the entire dataset is built from official PBP timestamps expanded into event windows. Only 20 of 152 games are manually checked, so systematic OCR or rule errors would corrupt all boundaries.
  • domain assumption Llama3-8B template-expanded PBP is acceptable ground-truth commentary.
    Section III-A: reference commentary is generated automatically, not written by human announcers. All commentary quality metrics are measured against this synthetic target.
  • domain assumption The frozen WASB ball detector provides reliable ball positions and confidence.
    Section IV-B.1: M1 and M2 anchor all visual extraction to ball motion. If ball detection fails during occlusions or fast plays, event and identity cues degrade.
  • domain assumption The game-level train, validation, and test split prevents broadcast-level leakage.
    Section III-B: the 121, 15, and 16 game split is stated, but no analysis of temporal or broadcast duplication across splits is provided.
  • domain assumption Event-type-specific boundary rules defined by three annotators generalize across all 152 games.
    Section III-A: boundaries are expanded from PBP timestamps using rules, and the 20-game manual check may not catch failures in rare event types.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NBA_Streaming: A Large-Scale Benchmark for Fine-Grained Basketball Commentary Generation in Continuous Streams." pith.science (2026). https://pith.science/paper/7KT5WND4

@misc{pith2026260809200,
  author       = {Pith},
  title        = {Pith review of: NBA_Streaming: A Large-Scale Benchmark for Fine-Grained Basketball Commentary Generation in Continuous Streams},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7KT5WND4}},
  note         = {Machine review of arXiv:2608.09200}
}
read the original abstract

Live basketball commentary generation requires determining when an event is sufficiently observable and describing it before subsequent events unfold. However, existing methods are primarily designed for pre-segmented clips or complete videos, making them unsuitable for continuous streams. Existing datasets also provide limited supervision for player identities, fine-grained actions, event attributes, and coherent event chains, restricting the factual richness of generated commentary. To address these limitations, we introduce NBA_Streaming, a large-scale benchmark for online fine-grained basketball commentary generation. It contains 307.5 hours of basketball broadcasts and approximately 35K temporally aligned events, with annotations of event boundaries, player identities, fine-grained actions, event chains, and natural-language commentary. By moving from isolated clips to continuous streams, NBA_Streaming enables unified evaluation of event localization, response reliability, factual grounding, and commentary quality under causal constraints. We further propose a causal two-stage framework that combines completion-first localization with ball-centric semantic grounding, enabling the system to identify complete events from observed streams and organize scene, event, identity, and action cues for commentary generation. Extensive experiments reveal the difficulty of NBA_Streaming, where existing baselines struggle with online timing, factual grounding, and fine-grained description. Our framework consistently improves over strong alternatives, while the remaining gap highlights NBA_Streaming as a valuable benchmark for streaming sports video understanding and generation. The code and data will be made publicly available upon acceptance.

Figures

Figures reproduced from arXiv: 2608.09200 by the authors.

Figure 1
Figure 1. Comparison between traditional clip-based basketball captioning and streaming basketball commentary. Unlike pre-segmented offline settings, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Construction pipeline of NBA Streaming. The pipeline integrates play￾by-play records and broadcast videos, aligns event timestamps with video time via OCR, annotates event boundaries and commentary, and merges causally related adjacent events into coherent event chains. Manual checking is then performed to ensure temporal coverage and semantic consistency. StreamMind [31] aim to perform online perception, memory upd… view at source ↗
Figure 3
Figure 3. Fine-grained action taxonomy of NBA Streaming. The taxonomy organizes basketball actions into 11 major action families and further divides them into 71 fine-grained action categories, covering diverse shooting, rebounding, defensive, foul, turnover, and other basketball actions. TABLE I COMPARISON OF NBA STREAMING WITH EXISTING VIDEO-LANGUAGE DATASETS. Dataset #Events #Sent. Dur. (h) Avg. Words Streaming Identity Se… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Overview of the proposed causal two-stage framework for streaming basketball commentary generation. Stage I performs completion-first localization [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 4
Figure 4. Figure 4: Overview of the proposed causal two-stage framework for streaming basketball commentary generation. Stage I performs completion-first localization [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison on four representative basketball events, highlighting event understanding, player identity recognition, and fine-grained action [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 2 canonical work pages

  1. [1]

    Vid2seq: Large-scale pretraining of a visual language model for dense video captioning,

    A. Yang, A. Nagrani, P. H. Seo, A. Miech, J. Pont-Tuset, I. Laptev, J. Sivic, and C. Schmid, “Vid2seq: Large-scale pretraining of a visual language model for dense video captioning,” in2023 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 10 714–10 726

    2023

  2. [2]

    Video recap: Recursive captioning of hour-long videos,

    M. M. Islam, N. Ho, X. Yang, T. Nagarajan, L. Torresani, and G. Bertasius, “Video recap: Recursive captioning of hour-long videos,”2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 18 198–18 208, 2024. [Online]. Available: https://api.semanticscholar.org/CorpusID:267759646

    2024

  3. [3]

    Show, think, and tell: Thought-augmented fine-tuning of large language models for video captioning,

    B. Kim, D. Hwang, S. Cho, Y . Jang, H. Lee, and M. Lee, “Show, think, and tell: Thought-augmented fine-tuning of large language models for video captioning,” in2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2024, pp. 1808–1817

    2024

  4. [4]

    Event- equalized dense video captioning,

    K. Wu, P. Li, J. Fu, Y . Li, Y . Wu, Y . Liu, J. Wang, and S. Zhou, “Event- equalized dense video captioning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025, pp. 8417–8427. 10

    work page 2025

  5. [5]

    Fine- grained video captioning for sports narrative,

    H. Yu, S. Cheng, B. Ni, M. Wang, J. Zhang, and X. Yang, “Fine- grained video captioning for sports narrative,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 6006–6015

    2018

  6. [6]

    Soccernet-caption: Dense video captioning for soccer broad- casts commentaries,

    H. Mkhallati, A. Cioppa, S. Giancola, B. Ghanem, and M. Van Droogen- broeck, “Soccernet-caption: Dense video captioning for soccer broad- casts commentaries,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 5074–5085

    2023

  7. [7]

    Goal: A challenging knowledge-grounded video captioning benchmark for real-time soccer commentary generation,

    J. Qi, J. Yu, T. Tu, K. Gao, Y . Xu, X. Guan, X. Wang, B. Xu, L. Hou, J. Liet al., “Goal: A challenging knowledge-grounded video captioning benchmark for real-time soccer commentary generation,” in Proceedings of the 32nd ACM international conference on information and knowledge management, 2023, pp. 5391–5395

    work page 2023

  8. [8]

    Matchtime: Towards automatic soccer game commentary generation,

    J. Rao, H. Wu, C. Liu, Y . Wang, and W. Xie, “Matchtime: Towards automatic soccer game commentary generation,” inProceedings of the 2024 Conference on Empirical Methods in Natural Language Process- ing, 2024, pp. 1671–1685

    2024

Show all 42 references
  1. [9]

    Player-centric multimodal prompt generation for large language model based identity-aware basketball video captioning,

    Z. Xi, H. Sun, Y . Wu, J. Yan, H. Zhang, L. Wu, L. Wang, and C. Chen, “Player-centric multimodal prompt generation for large language model based identity-aware basketball video captioning,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 2...

    2025

  2. [10]

    Knowledge guided entity-aware video captioning and a basketball benchmark,

    Z. Xi, G. Shi, X. Li, J. Yan, Z. Li, L. Wu, Z. Liu, and L. Wang, “Knowledge guided entity-aware video captioning and a basketball benchmark,”arXiv preprint arXiv:2401.13888, 2024

    2024 arXiv

  3. [11]

    Msr-vtt: A large video description dataset for bridging video and language,

    J. Xu, T. Mei, T. Yao, and Y . Rui, “Msr-vtt: A large video description dataset for bridging video and language,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 5288– 5296

    2016

  4. [12]

    A descriptive basketball highlight dataset for automatic commentary generation,

    B. Zhang, J. Gao, and Y . Yuan, “A descriptive basketball highlight dataset for automatic commentary generation,” inProceedings of the 32nd ACM international conference on multimedia, 2024, pp. 10 316– 10 325

    2024

  5. [13]

    Sports video analysis on large-scale data,

    D. Wu, H. Zhao, X. Bao, and R. P. Wildes, “Sports video analysis on large-scale data,” inEuropean conference on computer vision. Springer, 2022, pp. 19–36

    2022

  6. [14]

    A simple yet effective knowledge guided method for entity-aware video captioning on a basketball benchmark,

    Z. Xi, G. Shi, X. Li, J. Yan, Z. Li, L. Wu, Z. Liu, and L. Wang, “A simple yet effective knowledge guided method for entity-aware video captioning on a basketball benchmark,”Neurocomputing, vol. 619, p. 129177, 2025

    2025

  7. [15]

    Hierarchical representation network with auxiliary tasks for video captioning and video question answering,

    L. Gao, Y . Lei, P. Zeng, J. Song, M. Wang, and H. T. Shen, “Hierarchical representation network with auxiliary tasks for video captioning and video question answering,”IEEE Transactions on Image Processing, vol. 31, pp. 202–215, 2021

    2021

  8. [16]

    Concept-aware video captioning: Describing videos with effective prior information,

    B. Yang, M. Cao, and Y . Zou, “Concept-aware video captioning: Describing videos with effective prior information,”IEEE Transactions on Image Processing, vol. 32, pp. 5366–5378, 2023

    2023

  9. [17]

    Long-term action dependence-based hierarchical deep association for multi-athlete tracking in sports videos,

    L. Kong, D. Huang, and Y . Wang, “Long-term action dependence-based hierarchical deep association for multi-athlete tracking in sports videos,” IEEE Transactions on Image Processing, vol. 29, pp. 7957–7969, 2020

    2020

  10. [18]

    A novel identity-aware video captioning method with multi-perspective visual information for basketball,

    H. Li, H. Zhang, Z. Xi, H. Sun, and L. Wu, “A novel identity-aware video captioning method with multi-perspective visual information for basketball,”Engineering Applications of Artificial Intelligence, vol. 176, p. 114756, 2026

    2026

  11. [19]

    Egocentric online action segmentation via parametric context memory learning,

    X. Jiang, X. Xu, C. Liu, Z. Wang, J. Song, Z. Sun, A. Cichocki, and H. T. Shen, “Egocentric online action segmentation via parametric context memory learning,”IEEE Transactions on Image Processing, 2026

    2026

  12. [20]

    Fineaction: A fine- grained video dataset for temporal action localization,

    Y . Liu, L. Wang, Y . Wang, X. Ma, and Y . Qiao, “Fineaction: A fine- grained video dataset for temporal action localization,”IEEE transac- tions on image processing, vol. 31, pp. 6937–6950, 2022

    2022

  13. [21]

    Online temporal action local- ization with memory-augmented transformer,

    Y . Song, D. Kim, M. Cho, and S. Kwak, “Online temporal action local- ization with memory-augmented transformer,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 74–91

    2024

  14. [22]

    Hat: History- augmented anchor transformer for online temporal action localization,

    S. Reza, Y . Zhang, M. Moghaddam, and O. Camps, “Hat: History- augmented anchor transformer for online temporal action localization,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 205– 222

    2024

  15. [23]

    Onpoint: Offline-to-online multi-level distillation for point-supervised online temporal action localization,

    S. Reza, G. Jagatap, M. Moghaddam, O. Camps, and A. Fanelli, “Onpoint: Offline-to-online multi-level distillation for point-supervised online temporal action localization,”arXiv preprint arXiv:2607.00289, 2026

    2026 arXiv

  16. [24]

    Live video captioning,

    E. Blanco-Fern ´andez, C. Guti ´errez- ´Alvarez, N. Nasri, S. Maldonado- Basc´on, and R. J. L ´opez-Sastre, “Live video captioning,”Multimedia Tools and Applications, vol. 84, no. 35, pp. 44 863–44 895, 2025

    2025

  17. [25]

    Streaming dense video captioning,

    X. Zhou, A. Arnab, S. Buch, S. Yan, A. Myers, X. Xiong, A. Nagrani, and C. Schmid, “Streaming dense video captioning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 18 243–18 252

    2024

  18. [26]

    Livecc: Learning video llm with streaming speech transcription at scale,

    J. Chen, Z. Zeng, Y . Lin, W. Li, Z. Ma, and M. Z. Shou, “Livecc: Learning video llm with streaming speech transcription at scale,” inPro- ceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 29 083–29 095

    2025

  19. [27]

    Online video understanding: Ovbench and videochat-online,

    Z. Huang, X. Li, J. Li, J. Wang, X. Zeng, C. Liang, T. Wu, X. Chen, L. Li, and L. Wang, “Online video understanding: Ovbench and videochat-online,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 3328–3338

    2025

  20. [28]

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

    J. Chen, Z. Lv, S. Wu, K. Q. Lin, C. Song, D. Gao, J.-W. Liu, Z. Gao, D. Mao, and M. Z. Shou, “Videollm-online: Online video large language model for streaming video,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 18 407–18 418

    2024

  21. [29]

    Streaming videollms for real-time procedural video understanding,

    D. Chatterjee, E. Remelli, Y . Song, B. Tekin, A. Mittal, B. Bhatnagar, N. C. Camgoz, S. Hampali, E. Sauser, S. Maet al., “Streaming videollms for real-time procedural video understanding,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 22...

    2025

  22. [30]

    Flash-vstream: Efficient real-time understanding for long video streams,

    H. Zhang, Y . Wang, Y . Tang, Y . Liu, J. Feng, and X. Jin, “Flash-vstream: Efficient real-time understanding for long video streams,” inProceedings of the IEEE/CVF international conference on computer vision, 2025, pp. 21 059–21 069

    2025

  23. [31]

    Streammind: Unlocking full frame rate streaming video dia- logue through event-gated cognition,

    X. Ding, H. Wu, Y . Yang, S. Jiang, Q. Zhang, D. Bai, Z. Chen, and T. Cao, “Streammind: Unlocking full frame rate streaming video dia- logue through event-gated cognition,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 13 448–13 459

    2025

  24. [32]

    The llama 3 herd of models,

    A. Grattafiori, A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Vaughanet al., “The llama 3 herd of models,”arXiv preprint arXiv:2407.21783, 2024

    2024 arXiv

  25. [33]

    Widely applicable strong baseline for sports ball detection and tracking,

    S. Tarashima, M. A. Haq, Y . Wang, and N. Tagawa, “Widely applicable strong baseline for sports ball detection and tracking,”arXiv preprint arXiv:2311.05237, 2023

    2023 arXiv

  26. [34]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” inProceedings of the IEEE international conference on computer vision, 2017, pp. 2980–2988

    2017

  27. [35]

    Supervised contrastive learn- ing,

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y . Tian, P. Isola, A. Maschinot, C. Liu, and D. Krishnan, “Supervised contrastive learn- ing,”Advances in neural information processing systems, vol. 33, pp. 18 661–18 673, 2020

    2020

  28. [36]

    Class-balanced loss based on effective number of samples,

    Y . Cui, M. Jia, T.-Y . Lin, Y . Song, and S. Belongie, “Class-balanced loss based on effective number of samples,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 9268– 9277

    2019

  29. [37]

    Qwen3-vl technical report,

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

    2025 arXiv

  30. [38]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clarket al., “Learning transferable visual models from natural language supervision,” inInternational conference on machine learning. PmLR, 2021, pp. 8748–8763

    2021

  31. [39]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  32. [40]

    A systematic analysis of performance measures for classification tasks,

    M. Sokolova and G. Lapalme, “A systematic analysis of performance measures for classification tasks,”Information processing & manage- ment, vol. 45, no. 4, pp. 427–437, 2009

    2009

  33. [41]

    Cider: Consensus- based image description evaluation,

    R. Vedantam, C. Lawrence Zitnick, and D. Parikh, “Cider: Consensus- based image description evaluation,” inProceedings of the IEEE confer- ence on computer vision and pattern recognition, 2015, pp. 4566–4575

    2015

  34. [42]

    Bleu: a method for automatic evaluation of machine translation,

    K. Papineni, S. Roukos, T. Ward, and W.-J. Zhu, “Bleu: a method for automatic evaluation of machine translation,” inProceedings of the 40th annual meeting of the Association for Computational Linguistics, 2002, pp. 311–318

    2002

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.