Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation

Summary

Abstract Summary

Vision-Language-Action (VLA) models have recently achieved promising performance in robotic manipulation. However, existing benchmarks mainly evaluate generalization on static manipulation tasks and largely overlook dynamic interaction scenarios. To address this gap, we present ReflexBench, a benchmark for reaction-critical manipulation. ReflexBench contains six dynamic tasks and introduces an evaluation framework that decouples simulator stepping from robot control while supporting configurable latency under synchronous and asynchronous inference. Building upon ReflexBench, we propose ReflexVLA, an efficient VLA model designed for reaction-critical manipulation without large-scale robot-data pretraining. ReflexVLA enhances temporal reasoning through latent future prediction and multi-frame temporal fusion within the vision backbone, while reducing deployment latency through batched visual encoding and CUDA Graph replay. Experiments show that ReflexVLA consistently improves dynamic manipulation performance while maintaining competitive accuracy on standard static manipulation benchmarks, and real-world experiments further demonstrate its effectiveness under practical deployment conditions. Project website: https://reflexvla.github.io

Key Points

Vision-Language-Action (VLA) models have recently achieved promising performance in robotic manipulation.
However, existing benchmarks mainly evaluate generalization on static manipulation tasks and largely overlook dynamic interaction scenarios.
To address this gap, we present ReflexBench, a benchmark for reaction-critical manipulation
ReflexBench contains six dynamic tasks and introduces an evaluation framework that decouples simulator stepping from robot control while supporting configurable latency under synchronous and asynchronous inference.
Building upon ReflexBench, we propose ReflexVLA, an efficient VLA model designed for reaction-critical manipulation without large-scale robot-data pretraining.

ReflexBench

Recent robotic manipulation benchmarks evaluate VLA models across diverse tasks and environments [ 12 ] [ 13 ] [ 14 ] [ 15 ] , but largely overlook reaction-critical scenarios. Furthermore, most simulation benchmarks pause the environment during policy inference, ignoring inference latency and the resulting perception-execution delays.

Consequently, they fail to accurately assess VLA performance in dynamic real-world settings.

ReflexBench — Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation — Fig. 2 : Illustration of Four Inference Mechanisms. ReflexBench models the effect of policy inference latency by explicitly simulating the temporal gap between perception and action execution while allowing the latency to be adjusted in a controllable manner.

ReflexVLA

ReflexVLA is built on top of a compact vision-language-action architecture [ 10 ] . The visual encoder is a fused ViT backbone composed of DINOv2 [ 16 ] and SigLIP [ 17 ] at $224\times 224$ resolution. The language backbone is a Qwen2.5-0.5B [ 18 ] causal transformer.

Given a language instruction and multi-view RGB observations, the model inserts projected visual tokens after the beginning-of-sequence token and uses learnable action queries to produce an action chunk through a continuous regression head. As shown in Fig.

ReflexVLA — Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation — Fig. 3 : Overview of ReflexVLA. ReflexVLA is an efficient vision-language-action model for reaction-critical robotic manipulation, integrating latent future prediction, multi-frame temporal fusion, and inference latency optimization to enable fast and anticipatory decision-making.

Experiments

How do different factors influence the success rate of reaction-critical manipulation tasks? Q2: Does ReflexVLA achieve better performance on ReflexBench than existing VLA baselines? Q3: Does ReflexVLA still achieve competitive performance on standard static manipulation tasks? Q4: What is the contribution of each design component to the overall performance of ReflexVLA?

Q5: How well does ReflexVLA perform in real-world manipulation tasks? Baselines. We compare ReflexVLA with a diverse set of representative VLA models.

Experiments — Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation — Fig. 5 : Success rate matrices of SmolVLA under different chunk sizes and action horizons on two tasks. Chunk size is the policy’s training hyperparameter, while action horizon is the number of action steps executed during inference. The inference frequency is fixed at 30 Hz.

Sources and Demonstrations

Figures are reproduced from Chen et al. See the full paper for experimental details and the project page, when the paper links one, for demonstrations and videos.