AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

SwiftVLA: Unlocking Spatiotemporal Dynamics for Lightweight VLA Models at Minimal Overhead

Chaojun Ni, Cheng Chen, Xiaofeng Wang, Zheng Zhu, Wenzhao Zheng, Boyuan Wang, Tianrun Chen, Guosheng Zhao, Haoyun Li, Zhehao Dong, Qiang Zhang, Yun Ye and 3 more

CVPR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models · 22 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add 4D features as VLM input (ablation)SwiftVLA · RoboTwin 2.0average SR: 0.36 -> 0.40Improved
add Fusion Tokens with future trajectory supervision (ablation)SwiftVLA with 4D input · RoboTwin 2.0average SR: 0.40 -> 0.50Improved
add 4D feature masking during training (ablation)SwiftVLA (2D-only inference) · RoboTwin 2.0average SR: 0.02 -> 0.40Improved
add feature reconstruction to 4D masking (ablation)SwiftVLA (2D-only inference) · RoboTwin 2.0average SR: 0.40 -> 0.50Improved
add 2D feature masking (ablation)SwiftVLA (2D-only inference) · RoboTwin 2.0average SR: 0.50 -> 0.53Improved
randomize temporal cache size K in {3,4,5,6} during training instead of fixed (ablation)SwiftVLA · RoboTwin 2.0average SR: 0.52 -> 0.53Improved
increase fixed cache size from K=3 to K=6 (ablation)SwiftVLA · RoboTwin 2.0average SR: 0.47 -> 0.52Improved
use 4D input at inference in addition to mask-and-reconstruct trainingSwiftVLA · RoboTwin 2.0average SR: 0.53 -> 0.55No clear effect
use SwiftVLA instead of SmolVLAlightweight VLA · RoboTwin 2.0average SR: 0.29 -> 0.53Improved
use SwiftVLA instead of SmolVLAlightweight VLA · real-world tasksaverage score: 0.34 -> 0.80Improved
use SwiftVLA instead of pi0 on edge deviceVLA · Jetson Orininference time (s): 2.966 -> 0.167Improved
use SwiftVLA instead of OpenVLA-OFTVLA · LIBERO averagesuccess rate: 97.1 -> 94.7Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy