Beyond Attention Magnitude: Leveraging Inter-layer Rank Consistency for Efficient Vision-Language-Action Models
On the map: Vision-language-action models → Efficient VLA inference and adaptation
Abstract (arXiv)
Vision-Language-Action (VLA) models excel in robotic manipulation but suffer from significant inference latency due to processing dense visual tokens. Existing token reduction methods predominantly rely on attention magnitude as a static selection. In this work, we challenge this assumption, revealing that high-attention tokens are task-dependent and can even degrade policy performance. To address this, we introduce TIES (Tau-guided Inter-layer Efficient Selection), a dynamic framework guided by inter-layer token ranking consistency. By adaptively balancing attention magnitude with ranking consistency, TIES ensures robust token selection without requiring additional training. On the CogACT + SIMPLER benchmark, TIES improves average success rates by 6% while reducing token usage by 78%, and demonstrate strong generalization across diverse decoders and benchmarks.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 17 of its 31 references that are in the map. First posted on arXiv 26 Mar 2026.
Builds on
17 of 31 references.
Closest ancestors · 6
Same direction, newest first
- 2025SQAP-VLA: A Synergistic Quantization-Aware Pruning Framework for High-Performance Vision-Language-Action Models11 Sep
- 2025SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning6 Sep
- 2025SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration15 Jun
- 2025EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models11 Jun
- 2025Think Twice, Act Once: Token-Aware Compression and Action Reuse for Efficient Inference in Vision-Language-Action Models27 May
- 2024Quantization-Aware Imitation-Learning for Resource-Efficient Robotic Control2 Dec
Other references · 11
- 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
- 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
- 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
- 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
- 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
- 2023Vision Transformers Need Registerscited by 844
- 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
- 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
- 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
- 2025Stop Looking for Important Tokens in Multimodal Language Models: Duplication Matters Morecited by 149
- 2025A Survey on Efficient Vision-Language-Action Modelscited by 34
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2026Text-Vision Synergistic Token Caching: A Training-Free Framework for Efficient Vision-Language-Action Inference
- 2025The Better You Learn, The Smarter You Prune: Towards Efficient Vision-language-action Models via Differentiable Token Pruning
- 2026VLA-ACL: Action-Consistent Visual Token Pruning for Efficient Vision-Language-Action Models
- 2026Learning to Accelerate Vision-Language-Action Models through Adaptive Visual Token Caching
- 2026DTP: A Simple yet Effective Distracting Token Pruning Framework for Vision-Language Action Models
- 2025Action-aware Dynamic Pruning for Efficient Vision-Language-Action Manipulation
- 2026SAFE-Pruner: Semantic Attention-Guided Future-Aware Token Pruning for Efficient Vision-Language-Action Manipulation
- 2026BFA++: Hierarchical Best-Feature-Aware Token Prune for Multi-View Vision Language Action Model