pith. sign in

arxiv: 2605.29662 · v1 · pith:WJQPOGR7new · submitted 2026-05-28 · 💻 cs.CV · cs.RO

SAFE-Pruner: Semantic Attention-Guided Future-Aware Token Pruning for Efficient Vision-Language-Action Manipulation

Shilin Ma , Chubin Zhang , Changyuan Wang , Yuji Wang , Yue Wu , Zixuan Wang , Jingqi Tian , Zheng Zhu
show 1 more author
Yansong Tang
This is my paper

Pith reviewed 2026-06-29 08:38 UTC · model grok-4.3

classification 💻 cs.CV cs.RO
keywords vision-language-actiontoken pruningsemantic attentionefficient inferencerobotic manipulationattention consistencyfuture-aware pruningVLA models
0
0 comments X

The pith

Semantic attention consistency across steps lets SAFE-Pruner forecast deep-layer token importance and prune VLA models safely.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents SAFE-Pruner as a plug-and-play method that bases visual token removal on attention patterns expected in deeper layers rather than shallow cues alone. It rests on the observation that vision-language-action models keep their attention focused on the same semantic entity from one execution step to the next. Using this pattern to predict which tokens will still matter later avoids discarding information required downstream. An adaptive subtask split detects when attention jumps, improving the forecast. If the approach holds, real-time robotic control becomes faster while task success stays nearly unchanged.

Core claim

By identifying semantic attention consistency, the tendency that VLA models concentrate attention probability mass on the same semantic entity across execution steps, the method forecasts token saliency in deep layers to guide pruning, preventing premature removal of critical tokens and producing more stable acceleration than shallow-layer methods.

What carries the argument

semantic attention consistency, the observed concentration of attention on the same entity across steps, used to forecast deep-layer token saliency for pruning decisions

If this is right

  • Up to 1.89x inference speedup with success-rate degradation below 1.7 percent
  • Outperforms prior pruning methods by up to 1.9 percent in the same settings
  • Works as a plug-and-play addition to existing VLA models
  • Validated across simulation and real-world robotic manipulation tasks
  • Adaptive subtask division improves forecast accuracy when attention shifts abruptly

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same consistency pattern could appear in other sequential multimodal models, supporting similar pruning outside robotics
  • Lower compute could allow VLA policies to run on embedded hardware with tighter power budgets
  • Attention stability may vary with task type, so pruning gains could differ between navigation and fine manipulation
  • Forecasting from even earlier layers might compound the speedup if the consistency observation generalizes

Load-bearing premise

Attention remains concentrated on the same semantic entities from early layers through deep layers in a way that early cues reliably predict later needs.

What would settle it

A VLA task where early-layer attention forecasts cause removal of tokens required by later layers, producing success-rate drops above 5 percent on standard benchmarks.

Figures

Figures reproduced from arXiv: 2605.29662 by Changyuan Wang, Chubin Zhang, Jingqi Tian, Shilin Ma, Yansong Tang, Yue Wu, Yuji Wang, Zheng Zhu, Zixuan Wang.

Figure 1
Figure 1. Figure 1: Comparison of SAFE-Pruner and shallow-only token pruning meth [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Attention pattern differences and core token loss under early [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of semantic attention consistency and subtask tran [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The overall framework of SAFE-Pruner. Given a text prompt and a se￾quence of visual frames, multimodal tokens are produced by text and visual encoders. A saliency-based keyframe detection module identifies frames with significant saliency shifts. For keyframes, pruning is skipped; for non-keyframes, our future-aware token pruning module predicts late-stage token saliency from historical frames, fuses it wi… view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of the real-world experiment setup. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Real-world robotic task scenarios for evaluation. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
read the original abstract

Real-time inference of vision-language-action (VLA) models is essential for robotic control. While visual token pruning has shown strong potential for accelerating inference, most existing methods mainly base pruning decisions on shallow-layer cues and risk discarding visual information required by deep layers. To address this issue, we propose SAFE-Pruner, a plug-and-play pruning framework that incorporates attention cues of future layers into pruning decisions. Specifically, we identify semantic attention consistency, the tendency that VLA models concentrate their attention probability mass on the same semantic entity across execution steps. Based on this observation, we design a forward-looking strategy to forecast the token saliency in deep layers, which prevents the premature removal of critical tokens and leads to more stable acceleration. We further introduce an adaptive subtask division strategy to detect abrupt attention shifts, thereby improving forecasting accuracy and pruning reliability. Extensive experiments in simulation and real-world settings demonstrate that our method achieves up to 1.89x speedup with a minimal degradation in success rate of less than 1.7%, while outperforming state-of-the-art methods by up to 1.9%.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper introduces SAFE-Pruner, a plug-and-play token pruning framework for vision-language-action (VLA) models. It observes semantic attention consistency (concentration on the same semantic entity across steps) to forecast deep-layer token saliency and avoid premature pruning of critical tokens, augmented by an adaptive subtask division strategy to detect abrupt attention shifts. Experiments in simulation and real-world robotic manipulation report up to 1.89× speedup with <1.7% success-rate degradation while outperforming state-of-the-art pruning methods by up to 1.9%.

Significance. If the reported speedups and bounded performance loss hold under the stated conditions, the method offers a practical efficiency gain for real-time VLA inference in robotics. The forward-looking pruning and adaptive division are targeted responses to a known limitation of shallow-layer pruning; the plug-and-play design lowers the barrier to adoption.

major comments (1)
  1. [Abstract; §3 (method description of consistency and forecasting)] The central performance claims (1.89× speedup, <1.7% success-rate drop) rest on the reliability of semantic attention consistency for forecasting deep-layer saliency. The manuscript introduces adaptive subtask division to handle abrupt shifts but provides no explicit ablation or stress tests under distribution shift (e.g., added visual distractors) or extended task horizons; if consistency degrades, the saliency forecast becomes inaccurate and the degradation bound may not hold. This is load-bearing for the main empirical contribution.
minor comments (1)
  1. [Abstract] The abstract states quantitative results but does not name the specific VLA backbone, number of tokens pruned per layer, or exact baseline implementations; adding these details would improve reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the robustness of our core assumption. We agree that additional validation under distribution shifts and longer horizons would strengthen the empirical claims and will incorporate the requested experiments in the revision.

read point-by-point responses

  1. Referee: [Abstract; §3 (method description of consistency and forecasting)] The central performance claims (1.89× speedup, <1.7% success-rate drop) rest on the reliability of semantic attention consistency for forecasting deep-layer saliency. The manuscript introduces adaptive subtask division to handle abrupt shifts but provides no explicit ablation or stress tests under distribution shift (e.g., added visual distractors) or extended task horizons; if consistency degrades, the saliency forecast becomes inaccurate and the degradation bound may not hold. This is load-bearing for the main empirical contribution.

    Authors: We acknowledge that the manuscript does not include explicit ablations with added visual distractors or on substantially extended task horizons. Our current evaluation covers diverse simulation and real-world manipulation tasks that implicitly test varying conditions, and the adaptive subtask division is designed to detect attention shifts. However, to directly address the load-bearing concern, the revised version will add targeted stress-test ablations introducing visual distractors and evaluating longer-horizon sequences to confirm that the consistency-based forecasting remains reliable and the reported performance bounds hold. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical method grounded in external experiments

full rationale

The paper presents an empirical pruning framework that identifies semantic attention consistency via observation and uses it to design a forward-looking strategy, with performance validated through simulation and real-world experiments. No equations, fitted parameters renamed as predictions, self-citations, or self-definitional reductions appear in the provided text. The central claims reduce to measured speedups and success rates rather than to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that semantic attention consistency is a stable property of VLA models that can be exploited for forecasting; no free parameters or invented entities are described in the abstract.

axioms (1)
  • domain assumption Semantic attention consistency holds across execution steps in VLA models
    The pruning strategy is built directly on the stated tendency for models to concentrate attention probability mass on the same semantic entity.

pith-pipeline@v0.9.1-grok · 5756 in / 1262 out tokens · 37572 ms · 2026-06-29T08:38:16.928206+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 38 canonical work pages · 24 internal anchors

  1. [1]

    Alvar, S.R., Singh, G., Akbari, M., Zhang, Y.: Divprune: Diversity-based visual tokenpruningforlargemultimodalmodels.In:ProceedingsoftheComputerVision and Pattern Recognition Conference. pp. 9392–9401 (2025)

    2025

  2. [2]

    Qwen2.5-VL Technical Report

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., Zhong, H., Zhu, Y., Yang, M., Li, Z., Wan, J., Wang, P., Ding, W., Fu, Z., Xu, Y., Ye, J., Zhang, X., Xie, T., Cheng, Z., Zhang, H., Yang, Z., Xu, H., Lin, J.: Qwen2.5-vl technical report. arXiv preprint arXiv:2502.13923 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  3. [3]

    PaliGemma: A versatile 3B VLM for transfer

    Beyer, L., Steiner, A., Pinto, A.S., Kolesnikov, A., Wang, X., Salz, D., Neumann, M., Alabdulmohsin, I., Tschannen, M., Bugliarello, E., et al.: Paligemma: A ver- satile 3b vlm for transfer. arXiv preprint arXiv:2407.07726 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  4. [4]

    GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

    Bjorck, J., Castañeda, F., Cherniadev, N., Da, X., Ding, R., Fan, L., Fang, Y., Fox, D., Hu, F., Huang, S., et al.: Gr00t n1: An open foundation model for generalist humanoid robots. arXiv preprint arXiv:2503.14734 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    Black, K., Brown, N., Driess, D., Esmail, A., Equi, M., Finn, C., Fusai, N., Groom, L., Hausman, K., Ichter, B., et al.:π0: A vision-language-action flow model for general robot control. arXiv preprint arXiv:2410.24164 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  6. [6]

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

    Brohan, A., Brown, N., Carbajal, J., Chebotar, Y., Chen, X., Choromanski, K., Ding, T., Driess, D., Dubey, A., Finn, C., Florence, P., Fu, C., Arenas, M.G., Gopalakrishnan, K., Han, K., Hausman, K., Herzog, A., Hsu, J., Ichter, B., Irpan, A., Joshi, N., Julian, R., Kalashnikov, D., Kuang, Y., Leal, I., Lee, L., Lee, T.W.E., Levine, S., Lu, Y., Michalewski...

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  7. [7]

    RT-1: Robotics Transformer for Real-World Control at Scale

    Brohan, A., Brown, N., Carbajal, J., Chebotar, Y., Dabis, J., Finn, C., Gopalakr- ishnan, K., et al.: Rt-1: Robotics transformer for real-world control at scale. arXiv preprint arXiv:2212.06817 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  8. [8]

    In: European Conference on Computer Vision

    Chen, L., Zhao, H., Liu, T., Bai, S., Lin, J., Zhou, C., Chang, B.: 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. pp. 19–35. Springer (2024)

    2024

  9. [9]

    PaLI-X: On Scaling up a Multilingual Vision and Language Model

    Chen, X., Djolonga, J., Padlewski, P., Mustafa, B., Changpinyo, S., Wu, J., Ruiz, C.R., Goodman, S., Wang, X., Tay, Y., Shakeri, S., Dehghani, M., Salz, D., Lucic, M., Tschannen, M., Nagrani, A., Hu, H., Joshi, M., Pang, B., Montgomery, C., Pietrzyk, P., Ritter, M., Piergiovanni, A., Minderer, M., Pavetic, F., Waters, A., Li, 16 S. Ma et al. G., Alabdulmo...

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  10. [10]

    RLRC: Reinforcement Learning-based Recovery for Compressed Vision-Language-Action Models

    Chen, Y., Li, X.: Rlrc: Reinforcement learning-based recovery for compressed vision-language-action models. arXiv preprint arXiv:2506.17639 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  11. [11]

    The International Journal of Robotics Research44(10-11), 1684–1704 (2025)

    Chi,C.,Xu,Z.,Feng,S.,Cousineau,E.,Du,Y.,Burchfiel,B.,Tedrake,R.,Song,S.: Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research44(10-11), 1684–1704 (2025)

    2025

  12. [12]

    arXiv preprint arXiv:2402.08178 (2024)

    Choi, J.W., Yoon, Y., Ong, H., Kim, J., Jang, M.: Lota-bench: Bench- marking language-oriented task planners for embodied agents. arXiv preprint arXiv:2402.08178 (2024)

    work page arXiv 2024

  13. [13]

    Advances in neural information pro- cessing systems35, 16344–16359 (2022)

    Dao, T., Fu, D., Ermon, S., Rudra, A., Ré, C.: Flashattention: Fast and memory- efficient exact attention with io-awareness. Advances in neural information pro- cessing systems35, 16344–16359 (2022)

    2022

  14. [14]

    RoboNet: Large-Scale Multi-Robot Learning

    Dasari, S., Ebert, F., Tian, S., Nair, S., Bucher, B., Schmeckpeper, K., Singh, S., Levine, S., Finn, C.: Robonet: Large-scale multi-robot learning. arXiv preprint arXiv:1910.11215 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 1910

  15. [15]

    arXiv preprint arXiv:2510.09607 (2025)

    Dong, S., Fu, C., Gao, H., Zhang, Y.F., Yan, C., Wu, C., Liu, X., Shen, Y., Huo, J., Jiang, D., Cao, H., Gao, Y., Sun, X., He, R., Shan, C.: Vita-vla: Efficiently teaching vision-language models to act via action expert distillation. arXiv preprint arXiv:2510.09607 (2025)

    work page arXiv 2025

  16. [16]

    PaLM-E: An Embodied Multimodal Language Model

    Driess, D., Xia, F., Sajjadi, M.S.M., Lynch, C., Chowdhery, A., Ichter, B., Wahid, A., Tompson, J., Vuong, Q., Yu, T., Huang, W., Chebotar, Y., Sermanet, P., Duckworth, D., Levine, S., Vanhoucke, V., Hausman, K., Toussaint, M., Greff, K., Zeng, A., Mordatch, I., Florence, P.: Palm-e: An embodied multimodal language model. arXiv preprint arXiv:2303.03378 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  17. [17]

    Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasets

    Ebert, F., Yang, Y., Schmeckpeper, K., Bucher, B., Georgakis, G., Daniilidis, K., Finn, C., Levine, S.: Bridge data: Boosting generalization of robotic skills with cross-domain datasets. arXiv preprint arXiv:2109.13396 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  18. [18]

    arXiv preprint arXiv:2507.17141 (2025)

    Gao, G., Wang, J., Zuo, J., Jiang, J., Zhang, J., Zeng, X., Zhu, Y., Ma, L., Chen, K., Sheng, M., et al.: Towards human-level intelligence via human-like whole-body manipulation. arXiv preprint arXiv:2507.17141 (2025)

    work page arXiv 2025

  19. [19]

    $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization

    Intelligence, P., Black, K., Brown, N., Darpinian, J., Dhabalia, K., Driess, D., Esmail, A., Equi, M., Finn, C., Fusai, N., Galliker, M.Y., Ghosh, D., Groom, L., Hausman, K., Ichter, B., Jakubczak, S., Jones, T., Ke, L., LeBlanc, D., Levine, S., Li-Bell, A., Mothukuri, M., Nair, S., Pertsch, K., Ren, A.Z., Shi, L.X., Smith, L., Springenberg, J.T., Stachow...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  20. [20]

    arXiv preprint arXiv:2104.08212 (2021)

    Kalashnikov, D., Varley, J., Chebotar, Y., Swanson, B., Jonschkowski, R., Finn, C., Levine, S., Hausman, K.: Mt-opt: Continuous multi-task robotic reinforcement learning at scale. arXiv preprint arXiv:2104.08212 (2021)

    work page arXiv 2021

  21. [21]

    In: Forty-first International Conference on Machine Learning (2024)

    Karamcheti, S., Nair, S., Balakrishna, A., Liang, P., Kollar, T., Sadigh, D.: Pris- matic vlms: Investigating the design space of visually-conditioned language models. In: Forty-first International Conference on Machine Learning (2024)

    2024

  22. [22]

    Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success

    Kim, M.J., Finn, C., Liang, P.: Fine-tuning vision-language-action models: Opti- mizing speed and success. arXiv preprint arXiv:2502.19645 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  23. [23]

    OpenVLA: An Open-Source Vision-Language-Action Model

    Kim, M.J., Pertsch, K., Karamcheti, S., Xiao, T., Balakrishna, A., Nair, S., Rafailov, R., Foster, E., Lam, G., Sanketi, P., et al.: Openvla: An open-source vision-language-action model. arXiv preprint arXiv:2406.09246 (2024) SAFE-Pruner 17

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  24. [24]

    Efficient Memory Management for Large Language Model Serving with PagedAttention

    Kwon, W., Li, Z., Zhuang, S., Sheng, Y., Zheng, L., Yu, C.H., Gonzalez, J.E., Zhang, H., Stoica, I.: Efficient memory management for large language model serv- ing with pagedattention. arXiv preprint arXiv:2309.06180 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  25. [25]

    CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation

    Li, Q., Liang, Y., Wang, Z., Luo, L., Chen, X., Liao, M., Wei, F., Deng, Y., Xu, S., Zhang,Y.,etal.:Cogact:Afoundationalvision-language-actionmodelforsynergiz- ing cognition and action in robotic manipulation. arXiv preprint arXiv:2411.19650 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  26. [26]

    Evaluating Real-World Robot Manipulation Policies in Simulation

    Li, X., Hsu, K., Gu, J., Pertsch, K., Mees, O., Walke, H.R., Fu, C., Lunawat, I., Sieh, I., Kirmani, S., Levine, S., Wu, J., Finn, C., Su, H., Vuong, Q., Xiao, T.: Evaluating real-world robot manipulation policies in simulation. arXiv preprint arXiv:2405.05941 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    Advances in Neural Information Processing Systems36, 44776–44791 (2023)

    Liu, B., Zhu, Y., Gao, C., Feng, Y., Liu, Q., Zhu, Y., Stone, P.: Libero: Benchmark- ing knowledge transfer for lifelong robot learning. Advances in Neural Information Processing Systems36, 44776–44791 (2023)

    2023

  28. [28]

    Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference

    Liu, Z., Chen, Y., Cai, H., Lin, T., Yang, S., Liu, Z., Zhao, B.: Vla-pruner: Temporal-aware dual-level visual token pruning for efficient vision-language-action inference. arXiv preprint arXiv:2511.16449 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  29. [29]

    arXiv preprint arXiv:2602.02538 (2026)

    Lv, Z., Fan, Z., Tian, Q., Zhang, W., Zhuang, Y.: Enhancing post-training quan- tization via future activation awareness. arXiv preprint arXiv:2602.02538 (2026)

    work page arXiv 2026

  30. [30]

    IEEE Robotics and Automation Letters7(3), 7327–7334 (2022)

    Mees, O., Hermann, L., Rosete-Beas, E., Burgard, W.: Calvin: A benchmark for language-conditioned policy learning for long-horizon robot manipulation tasks. IEEE Robotics and Automation Letters7(3), 7327–7334 (2022)

    2022

  31. [31]

    In: International Conference on Learning Representations (ICLR) (2018)

    Micikevicius, P., Narang, S., Alben, J., Diamos, G., Elsen, E., Garcia, D., Ginsburg, B., Houston, M., Kuchaiev, O., Venkatesh, G., Wu, H.: Mixed precision training. In: International Conference on Learning Representations (ICLR) (2018)

    2018

  32. [32]

    In: 2024 IEEE International Conference on Robotics and Automation (ICRA)

    O’Neill, A., Rehman, A., Maddukuri, A., Gupta, A., Padalkar, A., Lee, A., Pooley, A., Gupta, A., Mandlekar, A., Jain, A., et al.: Open x-embodiment: Robotic learn- ing datasets and rt-x models: Open x-embodiment collaboration 0. In: 2024 IEEE International Conference on Robotics and Automation (ICRA). pp. 6892–6903. IEEE (2024)

    2024

  33. [33]

    arXiv preprint arXiv:2509.22093 (2025)

    Pei, X., Chen, Y., Xu, S., Wang, Y., Shi, Y., Xu, C.: Action-aware dynamic pruning for efficient vision-language-action manipulation. arXiv preprint arXiv:2509.22093 (2025)

    work page arXiv 2025

  34. [34]

    In: Meila, M., Zhang, T

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning trans- ferable visual models from natural language supervision. In: Meila, M., Zhang, T. (eds.) Proceedings of the 38th International Conference on Machine Learning. Proceedings of Machine Learning Res...

    2021

  35. [35]

    A Generalist Agent

    Reed, S., Zolna, K., Parisotto, E., Colmenarejo, S.G., Novikov, A., Barth-Maron, G., Gimenez, M., Sulsky, Y., Kay, J., Springenberg, J.T., Eccles, T., Bruce, J., Razavi, A., Edwards, A., Heess, N., Chen, Y., Hadsell, R., Vinyals, O., Bordbar, M., de Freitas, N.: A generalist agent. arXiv preprint arXiv:2205.06175 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  36. [36]

    Tian, J., Du, Y., Zhang, H., Wang, Y., Lee, I.N., Bai, X., Zhu, T., Niu, J., Tang, Y.: Ddavs: Disentangled audio semantics and delayed bidirectional alignment for audio-visual segmentation (2025),https://arxiv.org/abs/2512.20117

    work page arXiv 2025

  37. [37]

    Advances in Neural Information Processing Systems37, 114553–114573 (2024)

    Wang, C., Wang, Z., Xu, X., Tang, Y., Zhou, J., Lu, J.: Q-vlm: Post-training quantization for large vision-language models. Advances in Neural Information Processing Systems37, 114553–114573 (2024)

    2024

  38. [38]

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

    Wang, H., Nie, Y., Ye, Y., Wang, Y., Li, S., Yu, H., Lu, J., Huang, C.: Dynamic- vlm: Simple dynamic visual token compression for videollm. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 20812–20823 (2025) 18 S. Ma et al

    2025

  39. [39]

    SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning

    Wang, H., Xu, J., Pan, J., Zhou, Y., Dai, G.: Specprune-vla: Accelerating vision- language-action models via action-aware self-speculative pruning. arXiv preprint arXiv:2509.05614 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  40. [40]

    In: Che, W., Nabende, J., Shutova, E., Pilehvar, M.T

    Wang, Y., Zhang, H., Tian, J., Tang, Y.: Ponder & press: Advancing visual GUI agent towards general computer control. In: Che, W., Nabende, J., Shutova, E., Pilehvar, M.T. (eds.) Findings of the Association for Computational Linguistics: ACL 2025. pp. 1461–1473. Association for Computational Linguistics, Vienna, Austria (Jul 2025).https://doi.org/10.18653...

    work page doi:10.18653/v1/2025.findings- 2025

  41. [41]

    In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (2025)

    Xu, S., Wang, Y., Xia, C., Zhu, D., Huang, T., Xu, C.: Vla-cache: Efficient vision- language-action manipulation via adaptive token caching. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (2025)

    2025

  42. [42]

    arXiv preprint arXiv:2602.03782 (2026)

    Xu, Y., Yang, Y., Fan, Z., Liu, Y., Li, Y., Li, B., Zhang, Z.: Qvla: Not all channels are equal in vision-language-action model’s quantization. arXiv preprint arXiv:2602.03782 (2026)

    work page arXiv 2026

  43. [43]

    arXiv preprint arXiv:2506.10100 (2025)

    Yang, Y., Wang, Y., Wen, Z., Zhongwei, L., Zou, C., Zhang, Z., Wen, C., Zhang, L.: Efficientvla: Training-free acceleration and compression for vision-language-action models. arXiv preprint arXiv:2506.10100 (2025)

    work page arXiv 2025

  44. [44]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yang, Z., Wang, J., Tang, Y., Chen, K., Zhao, H., Torr, P.H.: Lavt: Language- aware vision transformer for referring image segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 18155– 18165 (2022)

    2022

  45. [45]

    Advances in Neural Information Processing Systems37, 56619–56643 (2024)

    Yue, Y., Wang, Y., Kang, B., Han, Y., Wang, S., Song, S., Feng, J., Huang, G.: Deer-vla:Dynamicinferenceofmultimodallargelanguagemodelsforefficientrobot execution. Advances in Neural Information Processing Systems37, 56619–56643 (2024)

    2024

  46. [46]

    arXiv preprint arXiv:2601.04061 (2026)

    Zhang, C., Wang, J., Gao, Z., Su, Y., Dai, T., Zhou, C., Lu, J., Tang, Y.: Clap: Contrastive latent action pretraining for learning vision-language-action models from human videos. arXiv preprint arXiv:2601.04061 (2026)

    work page arXiv 2026

  47. [47]

    arXiv preprint arXiv:2506.10967 (2025)

    Zhang, Q., Liu, M., Li, L., Lu, M., Zhang, Y., Pan, J., She, Q., Zhang, S.: Be- yond attention or similarity: Maximizing conditional diversity for token pruning in mllms. arXiv preprint arXiv:2506.10967 (2025)

    work page arXiv 2025

  48. [48]

    arXiv preprint arXiv:2503.20384 (2025)

    Zhang, R., Dong, M., Zhang, Y., Heng, L., Chi, X., Dai, G., Du, L., Du, Y., Zhang, S.: Mole-vla: Dynamic layer-skipping vision language action model via mixture-of- layers for efficient robot manipulation. arXiv preprint arXiv:2503.20384 (2025)

    work page arXiv 2025

  49. [49]

    SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference

    Zhang, Y., Fan, C.K., Ma, J., Zheng, W., Huang, T., Cheng, K., Gudovskiy, D., Okuno, T., Nakata, Y., Keutzer, K., et al.: Sparsevlm: Visual token sparsifica- tion for efficient vision-language model inference. arXiv preprint arXiv:2410.04417 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  50. [50]

    Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware

    Zhao, T.Z., Kumar, V., Levine, S., Finn, C.: Learning fine-grained bimanual ma- nipulation with low-cost hardware. arXiv preprint arXiv:2304.13705 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  51. [51]

    Segment Anything with Motion, Geometry, and Semantic Adaptation for Complex Nonlinear Visual Object Tracking

    Zhu, D., Wang, Y., Liu, Y., Tang, Y., Yu, B., Lu, J., Zhou, J.: Segment anything with motion, geometry, and semantic adaptation for complex nonlinear visual ob- ject tracking. arXiv preprint arXiv:2605.22538 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  52. [52]

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

    Zhu, T., Zhang, S., Shao, J., Tang, Y.: Kv-edit: Training-free image editing for precise background preservation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 16607–16617 (2025)

    2025

  53. [53]

    arXiv preprint arXiv:2512.18741 (2025) SAFE-Pruner 19

    Zhu, T., Zhang, S., Sun, Z., Tian, J., Tang, Y.: Memorize-and-generate: Towards long-term consistency in real-time video generation. arXiv preprint arXiv:2512.18741 (2025) SAFE-Pruner 19

    work page arXiv 2025

  54. [54]

    VARestorer: One-Step VAR Distillation for Real-World Image Super-Resolution

    Zhu, Y., Ma, S., Wang, H., Li, A., Jing, Y., Tang, Y., Chen, L., Lu, J., Zhou, J.: Varestorer: One-step var distillation for real-world image super-resolution. arXiv preprint arXiv:2604.21450 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  55. [55]

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

    Zhu, Y., Wang, H., Ma, S., Zhao, W., Tang, Y., Chen, L., Zhou, J.: Fade: Frequency-aware diffusion model factorization for video editing. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 28426–28435 (2025) SAFE-Pruner 1 A Detailed Implementation To validate the effectiveness and generalization of our method, we ev...

    2025