Artificial Intelligence · arXiv:2512.20276 v2 · submitted 23 Dec 2025
ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge
The paper says
“Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dynamic manipulation on edge hardware.”
from the abstract · picked out as the main claim by a classifier
Reviews
Reviews by people
No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their deployment in dynamic, real-world environments is severely hindered by high inference latency. While smooth robotic interaction requires control frequencies of 20-30 Hz, current VLA models typically operate at only 3-5 Hz on edge devices due to the memory-bound nature of autoregressive decoding. Existing optimizations often require extensive retraining or compromise model accuracy. To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge platforms. At the core of ActionFlow is a Cross-Request Pipelining strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. Furthermore, to support this scheduling, we propose a Cross-Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dynamic manipulation on edge hardware. Our work is available at https://github.com/Frankd35/Action-Flow.
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-11-03 · no language-model review
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.08 · states limitation 0.10 · human participants 0.03 · controlled comparison 0.28 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.57 · mentions code or data availability 0.98
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v2). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. | background 1.00 | 0.05 | 0.03 | asserted 1.00 |
| However, their deployment in dynamic, real-world environments is severely hindered by high inference latency. | background 0.86 | 0.09 | 0.02 | asserted 1.00 |
| While smooth robotic interaction requires control frequencies of 20-30 Hz, current VLA models typically operate at only 3-5 Hz on edge devices due to the memory-bound nature of autoregressive decoding. | background 0.91 | 0.16 | 0.03 | asserted 1.00 |
| Existing optimizations often require extensive retraining or compromise model accuracy. | background 0.52 | 0.04 | 0.02 | asserted 1.00 |
| To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge platforms. | contribution 1.00 | 0.82 | 0.06 | asserted 0.99 |
| At the core of ActionFlow is a Cross-Request Pipelining strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. | contribution 0.61 | 0.86 | 0.07 | asserted 1.00 |
| The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. | method 0.97 | 0.68 | 0.06 | asserted 1.00 |
| Furthermore, to support this scheduling, we propose a Cross-Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. | method 0.80 | 0.90 | 0.04 | asserted 1.00 |
| Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dynamic manipulation on edge hardware. | result 1.00 | 0.97 | 0.81 | asserted 1.00 |
| Our work is available at https://github.com/Frankd35/Action-Flow. | contribution 0.54 | 0.31 | 0.02 | asserted 0.76 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.99
What was examined, and what was not
Sources read
- arXiv listing abstract v2
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-03: 1 request, 7,587 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Cited by 1 Pith paper
Reviewed papers on Pith that cite this one. The full citation record
How the literature uses it
6 arXiv papers cite this in 8 sentences
Papers citing it · 4 sentences without a confident label
Cites it as background 3 papers
-
At the system level, ActionFlow [ref] and Real-time VLA [ref] lower on-device latency through operator fusion and pipelined scheduling, while VLASH [ref] overlaps inference with action execution to mitigate structural idle time.
EcoVLA: Energy-Efficient Device-Edge Co-Inference for Vision-Language-Action Models under Real-Time Constraints · VLA Deployment and Acceleration · 0.90
-
Existing approaches include adopting smaller VLM backbones [ref] , compressing LLM layers [ref] , accelerating action decoding [ref] , distilling diffusion models [ref] , pruning visual tokens [ref] , and applying optimization or quantization [ref] .
FASTER: Rethinking Real-Time Flow VLAs · Related Work · 0.93
-
ActionFlow [ref] pipelined prefill and decode phases for 2.55$×$ throughput on edge devices.
Reflex: Real-Time VLA Control through Streaming Inference · The Cache Reuse Dilemma · 0.82
Challenges and discussion
Know something about this paper? Leave a comment.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment