Pith. sign in

REVIEW 5 major objections 7 minor 46 references

EgoPrune: Efficient Token Pruning for Egomotion Video Reasoning in Embodied Agent

T0 review · 5 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read EgoPrune claims that a training-free, geometry-aware token-pruning pipeline—homography alignment, a 75% cosine-similarity cutoff, then prompt-balanced MMR selection—keeps over 99% of egomotion video-reasoning accuracy while cutting FLOPs…

desk verdict EgoPrune's accuracy story mostly holds; its headline efficiency claim is not yet backed by end-to-end measurement. read the letter →

arxiv 2507.15428 v1 pith:C4CQ5WU6 submitted 2025-07-21 cs.CV cs.AI

Jiaao Li , Kaiyuan Li , Chen Gao , Yong Li , Xinlei Chen This is my paper
classification cs.CVcs.AI
keywords egomotionvideoreasoningtraining-freetokenpruninghomographyalignmentvisualredundancyvision-languagemodelsembodiedAIedgedeploymentmaximalmarginalrelevance
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

EgoPrune is a training-free token-pruning method aimed at a specific problem: videos recorded by a moving agent, where every frame is a different viewpoint of the same environment. The paper claims that this egomotion redundancy can be removed with the camera's own geometry—estimate a homography between consecutive frames, warp one onto the other, and drop tokens whose aligned counterparts exceed a 75% cosine-similarity cutoff—followed by a Maximal Marginal Relevance selector that keeps tokens both relevant to the question and visually diverse. On the VSI-Bench and UrbanVideo-Bench egomotion benchmarks, retaining only 30–70% of visual tokens, the method reports average accuracy within a small fraction of the full-token baseline (35.16–35.43 vs. 35.45 on VSI-Bench; 45.66–46.25 vs. 46.23 on UrbanVideo-Bench) while lowering FLOPs, peak memory, and end-to-end latency. A deployment on a Jetson Orin NX 16GB edge device with an INT4-quantized 3B model is offered as evidence that the savings survive real embodied constraints. If the claim holds, spatially grounded video reasoning can run on-device for drones and mobile robots, and pruning methods for video no longer have to assume a fixed camera.

What carries the argument

The load-bearing object is the homography warp inside Perspective-Aware Redundancy Filtering (PARF): a $3\times3$ matrix $\mathbf{H}$ mapping each frame onto its successor, estimated from ORB matches via RANSAC, which turns viewpoint-shifted frames into a common coordinate system so that the 75% cosine-similarity threshold is a meaningful redundancy test. Because a single homography is exact only for planar scenes or pure rotation, the alignment is approximate in general egomotion, and the method's accuracy rests on that approximation being good enough. The second mechanism is the MMR token selector, which greedily maximizes $\lambda\cdot\mathrm{rel}(v_i)-(1-\lambda)\cdot\max_{j\in S}\mathrm{sim}(v_i,v_j)$, where relevance is the cosine similarity to the mean prompt-token vector and diversity is measured against a short window of earlier picks; $\lambda=0.5$ balances the two. Because neither stage reads attention scores, the pipeline composes with FlashAttention-2 and with quantized edge runtimes, which is what makes the on-device deployment possible.

What would settle it

Run EgoPrune at 30% token retention on egomotion videos where translation dominates and depth varies sharply—a handheld or drone camera moving straight through a cluttered room—and compare answer accuracy against the full-token baseline; the published gaps at 30% retention are about 0.3–0.6 points, so a drop of more than a point or two would falsify the alignment premise. A more direct check is to compute the per-token reprojection error of the RANSAC homography and verify that the tokens pruned as redundant are precisely the aligned ones, rather than a sizeable fraction of misaligned pixels.

Watch

Extended reading notes

Core claim

The discovery is that egomotion videos carry their redundancy in the camera's perspective, not in fixed pixel positions: after EmbodiedR's overlap-aware keyframe selection leaves roughly 50–60% overlap between consecutive frames, PARF estimates a homography from ORB keypoint matches filtered by FLANN and Lowe's ratio test, with RANSAC rejecting outliers; warping the previous frame onto the current one makes local token pairs comparable, and tokens with cosine similarity above 75% are declared redundant and dropped. The surviving tokens are then chosen by an MMR selector that scores each candidate by cosine similarity to the averaged prompt tokens, minus a penalty for similarity to already-selected tokens (with a window of ten selections keeping the cost at $O(n^2)$ in token count). The paper argues that this two-stage design answers the three failure modes of prior methods: fixed-position cosine similarity breaks under viewpoint shift, attention-based pruning collides with positional bias and FlashAttention, and diversity-only selection ignores the question being asked. On both benchmarks, pruned models at 70%, 50%, and 30% retention match or slightly beat the full-token average, with the clearest gains on perception-heavy tasks such as object appearance, size estimation, and counting.

Load-bearing premise

The load-bearing premise is that a single homography can align consecutive egomotion frames well enough for the 75% cosine-similarity cutoff to separate genuinely redundant tokens from informative ones; the homography assumption is exact only for planar scenes or purely rotational camera motion, so under real translation and depth variation the warp is wrong for many pixels and the filter can discard tokens the model still needs.

Editorial extensions

If this is right

  • At 70%, 50%, and 30% token retention on both benchmarks, EgoPrune's average accuracy stays within about a point of the full-token baseline (on VSI-Bench at 50% retention it is 35.43 vs. 35.45), so egomotion video redundancy can be largely removed without any fine-tuning.
  • FLOPs, peak memory, and end-to-end latency all decrease and scale more smoothly with input length than the DivPrune and PACT baselines, which is the property that makes long egomotion video usable in time-constrained embodied settings.
  • Ablations show the two stages are complementary: removing PARF hurts geometry-sensitive tasks (absolute distance, size estimation, route planning), while removing MMR hurts object-centric and directional tasks (object appearance, relative direction, cognitive map construction).
  • Because the method needs no attention scores, it can sit on top of FlashAttention-style kernels, and the Jetson Orin NX deployment with an INT4-quantized VILA-1.5 3B model shows the efficiency gains transfer to an edge runtime rather than existing only on server GPUs.
  • Accuracy staying flat from 70% down to 30% retention implies the token budget for egomotion reasoning can be chosen by the latency budget, since pruning strength does not trade away task accuracy in the tested range.

Reading between the lines

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

  • The single-homography model is the fragile point: for egomotion with strong translation through depth-varying scenes (walking past close objects, a drone flying toward buildings), one planar warp cannot align all pixels, so the 75% cutoff will misclassify some informative tokens as redundant; a depth-aware or multi-plane extension would be the natural stress test.
  • If the finding generalizes, the geometry-first principle—estimate the motion, prune in the aligned frame, then query-select—should transfer to other moving-camera settings such as driving videos, handheld AR capture, and robot manipulation, where the fixed-camera assumption of prior pruning methods fails just as it does in egomotion video.
  • The observation that pruned models sometimes beat the full-token baseline suggests the MMR stage acts as a soft query-grounded filter; a testable consequence is that adding the same MMR selection to other pruning methods (or to third-person video) would improve their accuracy at low retention independently of the egomotion setting.
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

5 major / 7 minor

Summary. The paper proposes EgoPrune, a training-free token pruning pipeline for egomotion video reasoning in vision-language models. It combines a keyframe selector adapted from EmbodiedR, a Perspective-Aware Redundancy Filtering (PARF) stage that uses homography-based alignment to drop temporally redundant tokens, and an MMR-based token selector that balances prompt relevance and intra-frame diversity. The method is evaluated on VSI-Bench and UrbanVideo-Bench with LLaVA-OneVision-7B against DivPrune and PACT at 70%, 50%, and 30% token retention, with additional efficiency measurements and an on-device deployment on a Jetson Orin NX 16GB. The abstract claims over 99% task accuracy preservation while significantly reducing FLOPs, memory, and latency.

Significance. If the efficiency claims hold, EgoPrune would be a useful contribution: it is training-free, geometry-aware, and plausibly deployable on embodied platforms. The appendix ablations of PARF and MMR are a strength, and the on-device evaluation adds practical evidence. However, the significance is currently tempered by unresolved accounting of pruning overhead, the absence of a keyframe-selector ablation, and the lack of statistical confidence in small accuracy differences. The core idea is promising and worth pursuing, but the empirical support needs to be tightened.

major comments (5)
  1. [Section 5.2.2 and Algorithm 1] The efficiency claim is not yet established. Figure 5 reports TFLOPs, latency, and memory with all methods retaining 50% of tokens, but the paper does not state whether the PARF (ORB+FLANN+RANSAC+warp) and MMR selection costs are included; DeepSpeed Profiler profiles model execution, not surrounding Python/pybind stages. If all methods feed the same number of tokens to the LLM, LLM-forward TFLOPs should be identical, so the reported differences must come from unequal final token counts or from un-reported pruning-stage costs. The paper needs an end-to-end accounting with and without pruning overhead, plus a clear statement of the actual token counts entering the LLM. In addition, Algorithm 1 line 8 computes max over all previously selected tokens j in S, not only the most recent w, so the claimed O(rwn^2) complexity with w=10 does not match the described procedure.
  2. [Section 4.2.1 and Section 5.3] The keyframe selector is never ablated. The pipeline's first stage is 'Embodied-R's overlap-aware keyframe selection', but no experiment compares EgoPrune with and without this stage or replaces it with uniform sampling. Consequently the reported gains over DivPrune and PACT cannot be attributed to PARF/MMR; they may largely stem from the keyframe selection. The appendix ablates PARF and MMR but holds the keyframe selector fixed. Please add a keyframe ablation and ensure baselines use the same frame-sampling policy.
  3. [Tables 1 and 2] The headline accuracy-preservation claim rests on very small differences. For example, VSI-Bench at 50% retention gives 35.43 for EgoPrune versus 35.45 for full tokens, and UrbanVideo-Bench at 30% gives 45.66 versus 46.23 for full tokens; the improvements over DivPrune and PACT are typically 0.2 to 1.0 points. No error bars, number of evaluation repeats, or significance tests are provided. Since the abstract's 'over 99%' claim is computed from these differences, the paper should report variance (e.g., bootstrap or multiple runs) and show that the ranking is stable.
  4. [Section 5.1.3] Excluding videos longer than 90 seconds removes the regime where token pruning matters most. At 1 FPS and 196 tokens per frame, 90 seconds yields about 17.6K visual tokens, which is moderate for a 7B VLM. The abstract's motivation is 'long, redundant video inputs', but the evaluation truncates long videos. Please report results on the full benchmarks, and/or analyze accuracy and efficiency as a function of video length, to justify the cutoff and the long-video claim.
  5. [Sections 3.2 and 4.2.1] PARF relies on a single homography per consecutive frame pair, which is exact only for planar scenes or pure rotation. Real egomotion includes translation with depth variation, so many tokens will be misaligned; the 75% cosine-similarity threshold could then discard informative tokens. The paper provides no alignment-quality metric or failure analysis. Please add quantitative evidence (e.g., fraction of frames where RANSAC inliers are below a threshold, or comparison to optical-flow or essential-matrix alignment) to show that misalignment does not undermine pruning decisions.
minor comments (7)
  1. [Figure 7 caption] The caption contains a typo: 'token retation rate' should be 'token retention rate'.
  2. [References] The RANSAC reference is given as Cantzler 1981; the standard reference is Fischler and Bolles (1981). Please correct this citation.
  3. [Throughout] The text repeatedly prints 'LLaV A' with a space; use 'LLaVA' consistently.
  4. [Section 5.1.3] The paper states three NVIDIA A6000 GPUs with 50GB, but the A6000 has 48GB memory; please correct or clarify the hardware configuration.
  5. [Algorithm 1] Algorithm 1 does not define the similarity function used in line 8; Section 4.2.2 defines sim after the algorithm, so the algorithm should refer explicitly to that definition.
  6. [Abstract and Section 5.2.1] The connection between 'over 99% of task accuracy' in the abstract and the numbers in Tables 1 and 2 should be made explicit, since the comparison is to the full-token baseline and not to a task-specific ceiling.
  7. [Section 4.2.1] The keyframe selector from EmbodiedR is not described sufficiently in this paper; a short algorithmic description or pseudocode would improve reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: self-citations provide components and benchmarks, but the central empirical claims are tested externally and do not reduce to prior equations or fitted values.

full rationale

EgoPrune's three components are assembled from cited prior work, including the authors' own EmbodiedR keyframe selector and the overlapping-author UrbanVideo-Bench. However, the paper's central claims—accuracy preservation and efficiency gains—are evaluated against external benchmarks (VSI-Bench and UrbanVideo-Bench) and against two baseline pruning methods (DivPrune and PACT) under matched token retention rates. The PARF step uses a fixed cosine-similarity threshold (75%) and a homography estimated from ORB/FLANN/RANSAC; the threshold is not fitted to the benchmark answers, and the ablation study shows that removing either PARF or MMR changes accuracy, indicating the components are not definitionally equivalent to the final result. The MMR selector is a deterministic greedy algorithm with a fixed λ = 0.5 in the main results; while λ is tuned on the same benchmarks, this is a hyperparameter choice rather than a fitted parameter renamed as a prediction. The cited EmbodiedR keyframe selector is a component, not a uniqueness theorem or a derivation of EgoPrune's performance, so the self-citation is not load-bearing in the sense of forcing the outcome. The efficiency measurements may undercount pruning-stage overhead, and the homography assumption may fail on non-planar scenes, but these are correctness risks, not circularity: the reported numbers would be false if the overhead were included, not true by construction. Overall, no step in the derivation chain reduces to its own inputs or to an unverified self-citation chain.

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

The method depends on several hand-set thresholds and inherited components; the most consequential are the homography-planarity assumption and the unablated keyframe selector.

free parameters (5)
  • Keyframe overlap threshold = not specified
    EmbodiedR's keyframe selector drops frames when visual overlap falls below a threshold; the threshold is inherited from prior work and not reported in this paper, yet it controls the number of frames that enter PARF.
  • PARF cosine similarity threshold = 0.75
    Tokens with cosine similarity above 75% are pruned as redundant; this threshold is stated without ablation or justification.
  • MMR trade-off lambda = 0.5
    Lambda balances prompt relevance and diversity; 0.5 is used for all main results, and the ablation shows sensitivity at 30% retention, indicating the choice matters.
  • MMR window size w = 10
    Diversity is computed against the 10 most recently selected tokens; the value is heuristic and not ablated.
  • Video length cutoff = 90 seconds
    Videos longer than 90 seconds are excluded to reflect deployment constraints; this is a data selection that changes the benchmark distribution.
assumptions (5)
  • domain assumption A single homography aligns consecutive egomotion frames (planar scene or purely rotational camera motion).
    PARF relies on this to establish token correspondences before cosine-similarity pruning; egomotion with translation and depth variation violates this.
  • domain assumption Retained tokens with distinct temporal embeddings still allow the model to perceive motion.
    Section 4.2.1 asserts that 3D RoPE temporal embeddings let the model understand motion without redundant tokens; this is not empirically verified.
  • domain assumption Adjacent keyframes from EmbodiedR retain 50%-60% visual overlap.
    Section 4.2.1 states this as the basis for the 75% similarity threshold; no measurement is shown for the benchmarks used.
  • domain assumption Cosine similarity to the average prompt token is a valid relevance measure for MMR.
    Follows CDPruner; assumes averaged prompt embeddings capture task relevance for spatial reasoning QA.
  • domain assumption The VSI-Bench and UrbanVideo-Bench QA pairs are valid measures of egomotion spatial reasoning.
    The paper relies on these benchmarks as ground truth for accuracy; one of them comes from overlapping authors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EgoPrune: Efficient Token Pruning for Egomotion Video Reasoning in Embodied Agent." pith.science (2026). https://pith.science/paper/C4CQ5WU6

@misc{pith2026250715428,
  author       = {Pith},
  title        = {Pith review of: EgoPrune: Efficient Token Pruning for Egomotion Video Reasoning in Embodied Agent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C4CQ5WU6}},
  note         = {Machine review of arXiv:2507.15428}
}
read the original abstract

Egomotion videos are first-person recordings where the view changes continuously due to the agent's movement. As they serve as the primary visual input for embodied AI agents, making egomotion video reasoning more efficient is therefore essential for real-world deployment. Recent advances in vision-language models have enabled strong multimodal reasoning capabilities, but their computational cost remains prohibitive for long, redundant video inputs. Existing token pruning methods, typically designed for third-person videos, fail to leverage the spatiotemporal continuity and motion constraints inherent in egomotion settings. To address this, we propose EgoPrune, a training-free token pruning method tailored for egomotion video reasoning. EgoPrune comprises three components: a keyframe selector adapted from EmbodiedR for temporally efficient sampling; Perspective-Aware Redundancy Filtering (PARF), which aligns visual tokens using perspective transformations and removes redundant tokens; and a Maximal Marginal Relevance (MMR)-based token selector that jointly considers visual-text relevance and intra-frame diversity. Experiments on two egomotion video benchmarks show that EgoPrune consistently outperforms prior training-free methods across various pruning ratios while significantly reducing FLOPs, memory usage, and latency. Moreover, we deploy EgoPrune on an embodied agent equipped with a Jetson Orin NX 16GB edge device, demonstrating its real-world efficiency and suitability for on-device egomotion video reasoning.

Figures

Figures reproduced from arXiv: 2507.15428 by the authors.

Figure 1
Figure 1. Identifying redundant tokens based on cosine [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of EgoPrune: Built on Embodied-R’s keyframe selection via overlap-based filtering, EgoPrune applies [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. By estimating a homography matrix between con [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Visualization of MMR Token Selector under dif [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Performance of EgoPrune across varying values [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: End-to-end latency (left) and peak GPU memory [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 1
Figure 1. Figure 1: This figure illustrates the efficiency comparison [PITH_FULL_IMAGE:figures/full_fig_p010_1.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 17 canonical work pages

  1. [1]

    L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al

    Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  2. [2]

    R.; Singh, G.; Akbari, M.; and Zhang, Y

    Alvar, S. R.; Singh, G.; Akbari, M.; and Zhang, Y. 2025. Divprune: Diversity-based visual token pruning for large multimodal models. In Proceedings of the Computer Vision and Pattern Recognition Conference, 9392--9401

    work page 2025

  3. [3]

    Y.; Rajbhandari, S.; Awan, A

    Aminabadi, R. Y.; Rajbhandari, S.; Awan, A. A.; Li, C.; Li, D.; Zheng, E.; Ruwase, O.; Smith, S.; Zhang, M.; Rasley, J.; et al. 2022. Deepspeed-inference: enabling efficient inference of transformer models at unprecedented scale. In SC22: International Conference for High Performance Computing, Networking, Storage and Analysis, 1--15. IEEE

    2022

  4. [4]

    Bai, J.; Bai, S.; Chu, Y.; Cui, Z.; Dang, K.; Deng, X.; Fan, Y.; Ge, W.; Han, Y.; Huang, F.; et al. 2023. Qwen technical report. arXiv preprint arXiv:2309.16609

    arXiv 2023

  5. [5]

    Cantzler, H. 1981. Random sample consensus (ransac). Institute for Perception, Action and Behaviour, Division of Informatics, University of Edinburgh, 3: 1--4

    work page 1981

  6. [6]

    Chen, L.; Zhao, H.; Liu, T.; Bai, S.; Lin, J.; Zhou, C.; and Chang, B. 2024. An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models. In European Conference on Computer Vision, 19--35. Springer

    work page 2024

  7. [7]

    Chung, C.; Park, Y.; Choi, S.; Ganbat, M.; and Choo, J. 2023. Shortcut-V2V: Compression Framework for Video-to-Video Translation based on Temporal Redundancy Reduction. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 7612--7622

    work page 2023

  8. [8]

    Dao, T.; Fu, D.; Ermon, S.; Rudra, A.; and R \'e , C. 2022. Flashattention: Fast and memory-efficient exact attention with io-awareness. Advances in neural information processing systems, 35: 16344--16359

    2022

Show all 46 references
  1. [9]

    Dhouib, M.; Buscaldi, D.; Vanier, S.; and Shabou, A. 2025. Pact: Pruning and clustering-based token reduction for faster visual language models. In Proceedings of the Computer Vision and Pattern Recognition Conference, 14582--14592

    2025

  2. [10]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

    2020 arXiv

  3. [11]

    Dutson, M.; Li, Y.; and Gupta, M. 2023. Eventful transformers: Leveraging temporal redundancy in vision transformers. In Proceedings of the IEEE/CVF international conference on computer vision, 16911--16923

    2023

  4. [12]

    J.; Jajal, P.; Davis, J

    Eliopoulos, N. J.; Jajal, P.; Davis, J. C.; Liu, G.; Thiravathukal, G. K.; and Lu, Y.-H. 2025. Pruning one more token is enough: Leveraging latency-workload non-linearities for vision transformers on the edge. In 2025 IEEE/CVF Winter Conference on Applications of Computer Visi...

    2025

  5. [13]

    Jin, Y.; Li, J.; Liu, Y.; Gu, T.; Wu, K.; Jiang, Z.; He, M.; Zhao, B.; Tan, X.; Gan, Z.; et al. 2024. Efficient multimodal large language models: A survey. arXiv preprint arXiv:2405.10739

    2024

  6. [14]

    Li, B.; Zhang, Y.; Guo, D.; Zhang, R.; Li, F.; Zhang, H.; Zhang, K.; Zhang, P.; Li, Y.; Liu, Z.; et al. 2024. Llava-onevision: Easy visual task transfer. arXiv preprint arXiv:2408.03326

    2024 arXiv

  7. [15]

    Li, K.; Chen, X.; Gao, C.; Li, Y.; and Chen, X. 2025 a . Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization. arXiv preprint arXiv:2505.22038

    2025

  8. [16]

    Li, Y.; Tang, C.; Zhuang, J.; Yang, Y.; Sun, G.; Li, W.; Ma, Z.; and Zhang, C. 2025 b . Improving llm video understanding with 16 frames per second. arXiv preprint arXiv:2503.13956

    2025 arXiv

  9. [17]

    Lin, J.; Yin, H.; Ping, W.; Molchanov, P.; Shoeybi, M.; and Han, S. 2024. Vila: On pre-training for visual language models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 26689--26699

    2024

  10. [18]

    Q.; Wang, J.; Soldan, M.; Wray, M.; Yan, R.; Xu, E

    Lin, K. Q.; Wang, J.; Soldan, M.; Wray, M.; Yan, R.; Xu, E. Z.; Gao, D.; Tu, R.-C.; Zhao, W.; Kong, W.; et al. 2022. Egocentric video-language pretraining. Advances in Neural Information Processing Systems, 35: 7575--7586

    2022

  11. [19]

    Lin, Z.; Lin, M.; Lin, L.; and Ji, R. 2025. Boosting multimodal large language models with visual tokens withdrawal for rapid inference. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, 5334--5342

    2025

  12. [20]

    Liu, H.; Li, C.; Wu, Q.; and Lee, Y. J. 2023. Visual instruction tuning. Advances in neural information processing systems, 36: 34892--34916

    2023

  13. [21]

    Maaz, M.; Rasheed, H.; Khan, S.; and Khan, F. S. 2023. Video-chatgpt: Towards detailed video understanding via large vision and language models. arXiv preprint arXiv:2306.05424

    2023 arXiv

  14. [22]

    Meng, Y.; Li, K.; Huang, C.; Gao, C.; Chen, X.; Li, Y.; and Zhang, X. 2025. PLPHP: Per-Layer Per-Head Vision Token Pruning for Efficient Large Vision-Language Models. arXiv preprint arXiv:2502.14504

    2025 arXiv

  15. [23]

    A.; Pistilli, F.; Alliegro, A.; and Averta, G

    Peirone, S. A.; Pistilli, F.; Alliegro, A.; and Averta, G. 2024. A backpack full of skills: Egocentric video understanding with diverse task perspectives. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18275--18285

    2024

  16. [24]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, 8748--8763. PmLR

    2021

  17. [25]

    Shi, L.; Zhang, H.; Yao, Y.; Li, Z.; and Zhao, H. 2024. Keep the cost down: A review on methods to optimize llm's kv-cache consumption. arXiv preprint arXiv:2407.18003

    2024 arXiv

  18. [26]

    L.; Papaioannou, I.; Eshghi, A.; Konstas, I.; and Lemon, O

    Suglia, A.; Greco, C.; Baker, K.; Part, J. L.; Papaioannou, I.; Eshghi, A.; Konstas, I.; and Lemon, O. 2024. Alanavlm: A multimodal embodied ai foundation model for egocentric video understanding. arXiv preprint arXiv:2406.13807

    2024 arXiv

  19. [27]

    Tang, X.; Qiu, J.; Xie, L.; Tian, Y.; Jiao, J.; and Ye, Q. 2025. Adaptive keyframe sampling for long video understanding. In Proceedings of the Computer Vision and Pattern Recognition Conference, 29118--29128

    2025

  20. [28]

    Tao, K.; Qin, C.; You, H.; Sui, Y.; and Wang, H. 2025. DyCoke: Dynamic Compression of Tokens for Fast Video Large Language Models. In Proceedings of the Computer Vision and Pattern Recognition Conference, 18992--19001

    2025

  21. [29]

    Tay, Y.; Dehghani, M.; Bahri, D.; and Metzler, D. 2022. Efficient transformers: A survey. ACM Computing Surveys, 55(6): 1--28

    2022

  22. [30]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  23. [31]

    Wen, Z.; Gao, Y.; Li, W.; He, C.; and Zhang, L. 2025. Token Pruning in Multimodal Large Language Models: Are We Solving the Right Problem? arXiv preprint arXiv:2502.11501

    2025 arXiv

  24. [32]

    H.; Wang, Y.; and Ma, J

    Wu, C. H.; Wang, Y.; and Ma, J. 2023. Maximal marginal relevance-based recommendation for product customisation. Enterprise Information Systems, 17(5): 1992018

    2023

  25. [33]

    Xing, L.; Huang, Q.; Dong, X.; Lu, J.; Zhang, P.; Zang, Y.; Cao, Y.; He, C.; Wang, J.; Wu, F.; et al. 2024. Pyramiddrop: Accelerating your large vision-language models via pyramid visual redundancy reduction. arXiv preprint arXiv:2410.17247

    2024 arXiv

  26. [34]

    Xu, S.; Wang, Y.; Xia, C.; Zhu, D.; Huang, T.; and Xu, C. 2025. VLA-Cache: Towards Efficient Vision-Language-Action Model via Adaptive Token Caching in Robotic Manipulation. arXiv preprint arXiv:2502.02175

    2025

  27. [35]

    W.; Han, R.; Fei-Fei, L.; and Xie, S

    Yang, J.; Yang, S.; Gupta, A. W.; Han, R.; Fei-Fei, L.; and Xie, S. 2025 a . Thinking in space: How multimodal large language models see, remember, and recall spaces. In Proceedings of the Computer Vision and Pattern Recognition Conference, 10632--10643

    2025

  28. [36]

    Yang, S.; Chen, Y.; Tian, Z.; Wang, C.; Li, J.; Yu, B.; and Jia, J. 2025 b . Visionzip: Longer is better but not necessary in vision language models. In Proceedings of the Computer Vision and Pattern Recognition Conference, 19792--19802

    2025

  29. [37]

    Yao, L.; Li, Y.; Wei, Y.; Li, L.; Ren, S.; Liu, Y.; Ouyang, K.; Wang, L.; Li, S.; Li, S.; et al. 2025. TimeChat-Online: 80\ arXiv preprint arXiv:2504.17343

    2025 arXiv

  30. [38]

    Zhai, X.; Mustafa, B.; Kolesnikov, A.; and Beyer, L. 2023. Sigmoid loss for language image pre-training. In Proceedings of the IEEE/CVF international conference on computer vision, 11975--11986

    2023

  31. [39]

    Zhang, H.; Li, X.; and Bing, L. 2023. Video-llama: An instruction-tuned audio-visual language model for video understanding. arXiv preprint arXiv:2306.02858

    2023 arXiv

  32. [40]

    A.; Hu, K.; Liu, S.; Zhang, Y.; Yang, J.; Li, C.; et al

    Zhang, K.; Li, B.; Zhang, P.; Pu, F.; Cahyono, J. A.; Hu, K.; Liu, S.; Zhang, Y.; Yang, J.; Li, C.; et al. 2024 a . Lmms-eval: Reality check on the evaluation of large multimodal models. arXiv preprint arXiv:2407.12772

    2024 arXiv

  33. [41]

    Zhang, Q.; Liu, M.; Li, L.; Lu, M.; Zhang, Y.; Pan, J.; She, Q.; and Zhang, S. 2025. Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs. arXiv preprint arXiv:2506.10967

    2025 arXiv

  34. [42]

    Zhang, Y.; Wu, J.; Li, W.; Li, B.; Ma, Z.; Liu, Z.; and Li, C. 2024 b . Video instruction tuning with synthetic data. arXiv preprint arXiv:2410.02713

    2024 arXiv

  35. [43]

    Zhao, B.; Fang, J.; Dai, Z.; Wang, Z.; Zha, J.; Zhang, W.; Gao, C.; Wang, Y.; Cui, J.; Chen, X.; et al. 2025 a . Urbanvideo-bench: Benchmarking vision-language models on embodied intelligence with video data in urban spaces. arXiv preprint arXiv:2503.06157

    2025

  36. [44]

    Zhao, B.; Wang, Z.; Fang, J.; Gao, C.; Man, F.; Cui, J.; Wang, X.; Chen, X.; Li, Y.; and Zhu, W. 2025 b . Embodied-R: Collaborative Framework for Activating Embodied Spatial Reasoning in Foundation Models via Reinforcement Learning. arXiv preprint arXiv:2504.12680

    2025 arXiv

  37. [45]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  38. [46]

    write newline

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

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.