mlsanityData to 7 Oct 2026

2D or 3D: Who Governs Salience in VLA Models? -- Tri-Stage Token Pruning Framework with Modality Salience Awareness

arXiv 2604.09244 ↗First posted 10 Apr 2026cs.MM cs.CV cs.ROcited by 0

On the map: Vision-language-action models → VLA architectures and generalization

also in 1 more
Abstract (arXiv)

Vision-Language-Action (VLA) models have emerged as the mainstream of embodied intelligence. Recent VLA models have expanded their input modalities from 2D-only to 2D+3D paradigms, forming multi-visual-modal VLA (MVLA) models. Despite achieving improved spatial perception, MVLA faces a greater acceleration demand due to the increased number of input tokens caused by modal expansion. Token pruning is an effective optimization methods tailored to MVLA models. However, existing token pruning schemes are designed for 2D-only VLA models, ignoring 2D/3D modality salience differences. In this paper, we follow the application process of multi-modal data in MVLA models and develop a tri-stage analysis to capture the discrepancy and dynamics of 2D/3D modality salience. Based on these, we propose a corresponding tri-stage token pruning framework for MVLA models to achieve optimal 2D/3D token selection and efficient pruning. Experiments show that our framework achieves up to a 2.55x inference speedup with minimal accuracy loss, while only costing 5.8% overhead. Our Code is coming soon.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in VLA architectures and generalization, one of 13 directions in Vision-language-action models. Builds on 30 of its 46 references that are in the map. First posted on arXiv 10 Apr 2026.

Builds on

30 of 46 references.

References · 30

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  2. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  3. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  4. 2023PaLM-E: An Embodied Multimodal Language Modelcited by 2,037
  5. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  6. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 507
  7. 2023Vision-Language Foundation Models as Effective Robot Imitatorscited by 344
  8. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  9. 2025Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policiescited by 90
  10. 2025EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Modelscited by 89
  11. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Cachingcited by 83
  12. 2024Diffusion-VLA: Generalizable and Interpretable Robot Foundation Model via Self-Generated Reasoningcited by 57
18 more
  1. 2025SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Accelerationcited by 56
  2. 2019PyRep: Bringing V-REP to Deep Robot Learningcited by 43
  3. 2025Running VLAs at Real-time Speedcited by 42
  4. 2025SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruningcited by 39
  5. 2025The Better You Learn, The Smarter You Prune: Towards Efficient Vision-language-action Models via Differentiable Token Pruningcited by 34
  6. 2025Think Twice, Act Once: Token-Aware Compression and Action Reuse for Efficient Inference in Vision-Language-Action Modelscited by 30
  7. 2026QVLA: Not All Channels Are Equal in Vision-Language-Action Model's Quantizationcited by 23
  8. 2025MLA: A Multisensory Language-Action Model for Multimodal Understanding and Forecasting in Robotic Manipulationcited by 22
  9. 2025EdgeVLA: Efficient Vision-Language-Action Modelscited by 19
  10. 2025Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inferencecited by 14
  11. 2026Any3D-VLA: Enhancing VLA Robustness via Diverse Point Cloudscited by 13
  12. 2026DyQ-VLA: Temporal-Dynamic-Aware Quantization for Embodied Vision-Language-Action Modelscited by 10
  13. 2026KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Modelscited by 9
  14. 2025Focusing on What Matters: Object-Agent-centric Tokenization for Vision Language Action modelscited by 6
  15. 2026HeiSD: Hybrid Speculative Decoding for Embodied Vision-Language-Action Models with Kinematic Awarenesscited by 6
  16. 2026RAPID: Redundancy-Aware and Compatibility-Optimal Edge-Cloud Partitioned Inference for Diverse VLA Modelscited by 4
  17. 2026RoboECC: Multi-Factor-Aware Edge-Cloud Collaborative Deployment for VLA Modelscited by 2
  18. 2026VLN-Cache: Enabling Token Caching for VLN Models with Visual/Semantic Dynamics Awarenesscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Revisiting Parameter Redundancy in Vision-Language-Action Models: Insights from VLM-to-VLA Adaptation
  2. 2026VLM4VLA: Revisiting Vision-Language-Models in Vision-Language-Action Models
  3. 2025UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent
  4. 2026VLA-Trace: Diagnosing Vision-Language-Action Models through Representation and Behavior Tracing
  5. 2025A Survey on Vision-Language-Action Models: An Action Tokenization Perspective
  6. 2025A Survey on Efficient Vision-Language-Action Models
  7. 2026VLANeXt: Recipes for Building Strong VLA Models
  8. 2025Vision-Language-Action (VLA) Models: Concepts, Progress, Applications and Challenges