DepthCache: Depth-Guided Training-Free Visual Token Merging for Vision-Language-Action Model Inference

• URL: http://arxiv.org/abs/2603.10469v1
• Date: Wed, 11 Mar 2026 06:40:44 GMT
• Title: DepthCache: Depth-Guided Training-Free Visual Token Merging for Vision-Language-Action Model Inference
• Abstract summary: Vision-Language-Action (VLA) models enable generalist robotic manipulation but suffer from high inference latency.<n>We present DepthCache, a training-free framework that leverages depth as a structural prior for visual token compression.<n>On the LIBERO benchmark, DepthCache achieves up to 1.28x inference speedup with less than 1% average success rate degradation.
• Score: 5.305950698447464
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision-Language-Action (VLA) models enable generalist robotic manipulation but suffer from high inference latency. This bottleneck stems from the massive number of visual tokens processed by large language backbones. Existing methods either prune or merge tokens uniformly, degrading the spatial reasoning essential for robotic control. We present DepthCache, a training-free framework that leverages depth as a structural prior for visual token compression. It partitions observations into depth-based regions and applies spatially differentiated merge ratios, preserving the near-field workspace while compressing the distant background. To exploit temporal redundancy, DepthCache distributes the merging process across consecutive frames, ensuring consistent representations while reducing per-step computation. A motion-adaptive pipeline further optimizes auxiliary view compression based on end-effector dynamics. The framework requires no model modification, generalizing across diverse VLA architectures. On the LIBERO benchmark, DepthCache achieves up to 1.28x inference speedup with less than 1% average success rate degradation across three VLA models (pi_0.5, OpenVLA, GR00T), whereas pruning and merging baselines incur 4--24% degradation at comparable compression. Real-world experiments on a physical manipulator demonstrate that DepthCache enables faster task throughput and more responsive closed-loop control in latency-sensitive scenarios.

Related papers

• VisCache: Visual KV Cache Pruning for Efficient Vision Large Language Model Inference [17.932004205808486]
  VisCache is a framework for coarse-to-fine textbfVisual KV textbfCache pruning without training.<n>PruneKV is a surgical KV compression algorithm tailored to the attention dynamics of Vision Large Language Models.<n>VisCache consistently outperforms existing baselines.
  arXiv  Detail & Related papers  (2026-08-25T04:52:05Z)
• Reducing Temporal Redundancy for Efficient Vision-Language-Action Inference [19.683948028719808]
  Vision-Language-Action (VLA) models exhibit strong generalization for robotic manipulation, yet their high inference latency limits real time deployment.<n>We propose a system level acceleration strategy that reduces computation in both perception and action generation.<n> Experiments on Libero, RobotWin, and Real Robot Platforms demonstrate over 2 times speedup while maintaining high performance.
  arXiv  Detail & Related papers  (2026-07-14T02:48:31Z)
• Depth Adaptive Efficient Visual Autoregressive Modeling [18.16894544431278]
  Visual Autoregressive ( VAR) modeling inefficiently applies a fixed computational depth to each position when generating high-resolution images.<n>We propose a paradigm shift from pruning entire tokens to adaptively allocating per-token computational depth.
  arXiv  Detail & Related papers  (2026-04-19T06:59:34Z)
• WorldCache: Content-Aware Caching for Accelerated Video World Models [50.7543797435026]
  We present textbfWorldCache, a Perception-Constrained Dynamic CaCache framework.<n>WorldCache improves both when and how to reuse features.<n>On Cosmos-2.5-2B evaluated on PAI-Bench, WorldCache achieves textbf$2.3times$ inference speedup while preserving textbf99.4% of baseline quality.
  arXiv  Detail & Related papers  (2026-03-23T17:59:54Z)
• PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cache [61.57938553036056]
  We introduce PackCache, a training-free KV-cache management method that compacts the KV cache through three coordinated mechanisms.<n>In terms of efficiency, PackCache accelerates end-to-end generation by 1.7-2.2x on 48-frame long sequences.
  arXiv  Detail & Related papers  (2026-01-07T19:51:06Z)
• OmniSAT: Compact Action Token, Faster Auto Regression [70.70037017501357]
  We introduce an Omni Swift Action Tokenizer, which learns a compact, transferable action representation.<n>The resulting discrete tokenization shortens the training sequence by 6.8$times$, and lowers the target entropy.
  arXiv  Detail & Related papers  (2025-10-08T03:55:24Z)
• Sliding Window Attention for Learned Video Compression [67.57073402826292]
  This work introduces 3D Sliding Window Attention (SWA), a patchless form of local attention.<n>Our method significantly improves rate-distortion performance, achieving Bjorntegaard Delta-rate savings of up to 18.6 %.
  arXiv  Detail & Related papers  (2025-10-04T20:11:43Z)
• FastCache: Fast Caching for Diffusion Transformer Through Learnable Linear Approximation [43.83288560196838]
  Diffusion Transformers (DiT) are powerful generative models but remain computationally intensive due to their iterative structure and deep transformer stacks.<n>FastCache is a hidden-state-level caching and compression framework that accelerates DiT inference.<n> Empirical evaluations across multiple DiT variants demonstrate substantial reductions in latency and memory usage.
  arXiv  Detail & Related papers  (2025-05-26T05:58:49Z)
• dKV-Cache: The Cache for Diffusion Language Models [53.85291644298835]
  Diffusion Language Models (DLMs) have been seen as a promising competitor for autoregressive language models.<n>We propose a KV-cache-like mechanism, delayed KV-Cache, for the denoising process of DLMs.<n>Our approach is motivated by the observation that different tokens have distinct representation dynamics throughout the diffusion process.
  arXiv  Detail & Related papers  (2025-05-21T17:32:10Z)
• VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching [23.52474883720957]
  Vision-Language-Action (VLA) models have demonstrated strong multi-modal reasoning capabilities, enabling direct action generation from visual perception and language instructions.<n>This paper introduces VLA-Cache, a training-free inference acceleration method that reduces computational overhead by adaptively caching and reusing static visual tokens across frames.
  arXiv  Detail & Related papers  (2025-02-04T09:48:14Z)
• Dynamic Frame Interpolation in Wavelet Domain [57.25341639095404]
  Video frame is an important low-level computation vision task, which can increase frame rate for more fluent visual experience. Existing methods have achieved great success by employing advanced motion models and synthesis networks. WaveletVFI can reduce computation up to 40% while maintaining similar accuracy, making it perform more efficiently against other state-of-the-arts.
  arXiv  Detail & Related papers  (2023-09-07T06:41:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.