Awesome-Dual-System-VLA

November 21, 2025 · View on GitHub

To address the challenges in vision-language-action (VLA) models, such as the difficulty of achieving efficient real-time performance, the high cost of pre-training, and the complexity of end-to-end fine-tuning on embodied data due to domain shift and catastrophic forgetting, Dual-System VLA models were introduced.

The development and architectural details of Dual-System VLA models are discussed in the paper.

This repository will be continuously updated, and we warmly welcome contributions from the community. If you have papers, projects, or resources that are not yet included, please feel free to submit them via a pull request or open an issue for discussion.

Current Results

CALVIN ABC→D

Method	1	2	3	4	5	Avg. Len.
Single-System
OpenVLA	91.3	77.8	62.0	52.1	43.5	3.27
UniVLA	95.5	85.8	75.4	66.9	56.5	3.80
Seer	94.4	87.2	79.9	72.2	64.3	3.98
Dual-System
LCB	73.6	50.2	28.5	16.0	9.9	1.78
RationalVLA	74.3	58.3	42.3	30.0	20.7	2.26
Robodual	94.4	82.7	72.1	62.4	54.4	3.66
OpenHelix	97.1	91.4	82.8	72.6	64.1	4.08

LIBERO

Method	LIBERO-Spatial	LIBERO-Object	LIBERO-Goal	LIBERO-Long	Avg.
Single-System
OpenVLA	84.7	88.4	79.2	53.7	76.5
π₀	96.8	98.8	95.8	85.2	94.2
OpenVLA-OFT	97.6	98.4	97.9	94.5	97.1
GR00T N1	94.4	97.6	90.6	93.9	93.9
UniVLA	96.5	96.8	95.6	92.0	95.2
Seer	-	-	-	87.7	-
Dual-System
DexVLA	97.2	99.1	95.6	-	-
Hume	98.6	99.8	99.4	98.6	98.6

✅ Dual-System VLA

Robot Manipulation

Title	Venue	Date	Code
Galaxea Open-World Dataset and G0 Dual-System VLA Model	arXiv	2025-08-30	Github
TriVLA: A Unified Triple-System-Based Unified Vision-Language-Action Model for General Robot Control	arXiv	2025-07-01	Project
RationalVLA: A Rational Vision-Language-Action Model with Dual System	arXiv	2025-06-12	Project
Fast-in-Slow: A Dual-System Foundation Model Unifying Fast Manipulation within Slow Reasoning	arXiv	2025-06-02	Github
Hume: Introducing System-2 Thinking in Visual-Language-Action Model	arXiv	2025-05-27	Github
OpenHelix: A Short Survey, Empirical Analysis, and Open-Source Dual-System VLA Model for Robotic Manipulation	arXiv	2025-05-06	Github
DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Control	arXiv	2025-02-09	Github
RoboDual: Towards Synergistic, Generalized, and Efficient Dual-System for Robotic Manipulation	arXiv	2024-10-10	Github
DP-VLA: A Dual Process VLA: Efficient Robotic Manipulation Leveraging VLM	CoRL 2024	2024-10-21	-
HiRT: Enhancing Robotic Control with Hierarchical Robot Transformers	CoRL 2024	2024-09-12	-
LCB: From LLMs to Actions: Latent Codes as Bridges in Hierarchical Robot Control	IROS 2024	2024-05-08	Project

Humanoid Robot

Title	Venue	Date	Code
Helix: A Vision-Language-Action Model for Generalist Humanoid Control	-	-	Porject

❌ Not a Dual-System VLA

Robot Manipulation

Title	Venue	Date	Code
OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoning	arXiv	2025-05-17	Github
NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Tasks	arXiv	2025-04-28	Github
π_0.5: a Vision-Language-Action Model with Open-World Generalization	arXiv	2025-04-22	Project
π₀: A Vision-Language-Action Flow Model for General Robot Control	arXiv	2024-10-31	Project
PIVOT-R: Primitive-Driven Waypoint-Aware World Model for Robotic Manipulation	NeurIPS 2024	2024-10-14	Github
MResT: Multi-Resolution Sensing for Real-Time Control with Vision-Language Models	CoRL 2023	2024-01-25	Github

Humanoid Robot

Title	Venue	Date	Code
GR00T N1: An Open Foundation Model for Generalist Humanoid Robots	arXiv	2025-03-18	Porject