ST-VLA: Enabling 4D-Aware SpatioTemporal Understanding
for General Robot Manipulation

You Wu* 1, Zixuan Chen* 2, Cunxu Ou2, Wenxuan Wang2, Wenbo Huang3, Lin Cao4, Yangtao Chen1, Weichao Qiu5, Xingyue Quan5, Jieqi Shi2, Jing Huo1, Yang Gao2
1School of Computer Science, Nanjing University, 2School of Intelligence Science and Technology, Nanjing University, 3School of Electronic Science and Engineering, Nanjing University, 4School of Instrument Science and Engineering, Southeast University, 5Noah's Ark Lab, Huawei
*Equal contribution. ✉ Correspondence to: You Wu <you@smail.nju.edu.cn>
ST-VLA Overview

ST-VLM bridges the semantic-physical gap via unified 3D-4D spatio-temporal representations. (Left) Existing 2D-based VLMs face geometric ambiguity and temporal inconsistency due to the semantic-physical mismatch. (Right) Our ST-VLA utilizes unified 3D-4D representations with explicit trajectories and smooth spatial masks, ensuring robust long-horizon manipulation.

Abstract

Robotic manipulation in open-world environments requires reasoning across semantics, geometry, and long-horizon action dynamics. Existing Vision-Language-Action (VLA) frameworks typically use 2D representations to connect high-level reasoning with low-level control, but lack depth awareness and temporal consistency, limiting robustness in complex 3D scenes. We propose ST-VLA, a hierarchical VLA framework using a unified 3D-4D representation to bridge perception and action. ST-VLA converts 2D guidance into 3D trajectories and generates smooth spatial masks that capture 4D spatio-temporal context, providing a stable interface between semantic reasoning and continuous control. To enable effective learning of such representations, we introduce ST-Human, a large-scale human manipulation dataset with 14 tasks and 300k episodes, annotated with 2D, 3D, and 4D supervision via a semi-automated pipeline. Using ST-Human, we train ST-VLM, a spatio-temporal vision-language model that generates spatially grounded and temporally coherent 3D representations to guide policy execution. The smooth spatial masks focus on task-relevant geometry and stabilize latent representations, enabling online replanning and long-horizon reasoning. Experiments on RLBench and real-world manipulation tasks show that ST-VLA significantly outperforms state-of-the-art baselines, improving zero-shot success rates by 44.6% and 30.3%. These results demonstrate that offloading spatio-temporal reasoning to VLMs with unified 3D-4D representations substantially improves robustness and generalization for open-world robotic manipulation.

The ST-VLA Pipeline

ST-VLA Pipeline

The ST-VLA Pipeline. Given a global instruction and an RGB-D observation, the high-level ST-VLM generates sub-instructions and 2D trajectories. These are lifted to 3D and fused with SAM2 masks to form a unified 3D-4D representation, which conditions the low-level 3D policy for continuous action execution. Guidance is refreshed every H steps for replanning and robustness to disturbances.


ST-Human Dataset & Task Generation

ST-Human Dataset

Overview of the ST-Human Dataset Construction and Unified 2D-3D-4D Task Generation.

Experiments & Results

Real-World Evaluation

Setup & Tasks: (Left) Franka Emika Panda hardware setup. (Right) Qualitative results showing: (1) zero-shot generalization; (2) distractor robustness; (3) long-horizon chaining.

Real World Setup

Performance: Real-world zero-shot generalization results.

Real World Results

Quantitative Comparisons

Table 1: VLM Performance Comparison on Selected 2D, 3D, and 4D Benchmarks. Best results are bolded.

Table 2: Results on Simulated Robot Manipulation Tasks. Background colors highlight ST-VLA variants.

Table 3: Success rates on long-horizon push-button tasks across three seeds, evaluated ST-VLA(3DFA) on seen buttons and unseen multi-step sequences.

Simulation Demos (3DFA (Baseline) vs our ST-VLA models)

evaluated on

3DFA (Baseline) -

Baseline RGB

RGB

Baseline Depth

Depth

ST-VLA (Ours) -

Ours RGB

RGB

Ours Depth

Depth

Loading demos...

Real-World Demos