Skip to content
AI News HubLIVE
Original source2 min read

EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control

Summary

EvoScene-VLA is a new vision-language-action (VLA) policy that maintains a persistent action-updated scene prior across control chunks, addressing scene changes caused by robot actions. It achieves 89.1% success rate on RoboTwin tasks and outperforms baselines on a real robot.

SourcearXiv RoboticsAuthor: Chushan Zhang, Ruihan Lu, Jinguang Tong, Xuesong Li, Yikai Wang, Hongdong Li
EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[2605.21862] EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control

[Submitted on 21 May 2026]

Title:EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control

View a PDF of the paper titled EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control, by Chushan Zhang and 5 other authors

View PDF HTML (experimental)

Abstract:Chunked vision-language-action (VLA) policies predict multi-step robot controls, conditioning each update on the current visual observation alone. Yet robot actions cause contact, occlusion, and object motion, and the geometry that later decisions depend on can change before the next visual update arrives. Spatial VLAs improve current-frame geometry. Temporal VLAs aggregate past frames. Neither maintains an action-updated scene prior across chunks. We argue for a persistent action-updated scene state across control calls, and introduce EvoScene-VLA. Its recurrent scene prefix carries a geometry-aware scene state across chunks. At each vision-language model (VLM) call, the VLM combines scene information from the current observation with the action-updated prior from the previous chunk; the action decoder outputs both the next action chunk and a compact scene update. This update becomes the next prior, which the VLM corrects against the new observation when the next call arrives. Each control call therefore starts from a scene prior that reflects both recent actions and fresh visual evidence. During training, \textbf{Scene Predictor} supplies future scene-token targets, and Geometric Anchor aligns scene slots with frozen depth and 3D teachers. We discard both modules at deployment. On 31 RoboTwin tasks, EvoScene-VLA raises average success from 87.2% to 89.1% in fixed evaluation and from 86.1% to 88.5% in randomized evaluation. On the Galaxea R1-Lite real robot, EvoScene-VLA outperforms all baselines.

Subjects:

Robotics (cs.RO); Artificial Intelligence (cs.AI)

Cite as: arXiv:2605.21862 [cs.RO]

(or arXiv:2605.21862v1 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2605.21862

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Jinguang Tong [view email] [v1] Thu, 21 May 2026 01:19:17 UTC (14,432 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control, by Chushan Zhang and 5 other authors

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.RO

new | recent | 2026-05

Change to browse by:

cs cs.AI

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • EvoScene-VLA uses a recurrent scene prefix to carry a geometry-aware scene state across chunks, combining current observations with action-updated priors.
  • It raises average success from 87.2% to 89.1% in fixed evaluation and from 86.1% to 88.5% in randomized evaluation on 31 RoboTwin tasks.
  • Training uses Scene Predictor and Geometric Anchor modules, which are discarded at deployment for efficiency.

Highlights and analysis are generated automatically and may contain errors. Check the original source.