Training Vision-Language-Action Models with Dense Embodied Chain-of-Thought Supervision

• URL: http://arxiv.org/abs/2606.30552v2
• Date: Wed, 01 Jul 2026 17:07:04 GMT
• Title: Training Vision-Language-Action Models with Dense Embodied Chain-of-Thought Supervision
• Abstract summary: Cross-embodiment transfer in vision-language-action (VLA) models remains challenging.<n>We present ZR-0, a 2.6 billion parameter end-to-end VLA model that uses dense Embodied Chain-of-Reason (ECoT) supervision to align cross-embodiment representations.
• Score: 26.793499191581315
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Cross-embodiment transfer in vision-language-action (VLA) models remains challenging because low-level state and action spaces differ fundamentally across robot platforms. We observe that the high-level cognitive process underlying manipulation, including scene perception, object identification, task planning, and sub-task decomposition, is largely shared across embodiments. Based on this observation, we present ZR-0, a 2.6 billion parameter end-to-end VLA model that uses dense Embodied Chain-of-Thought (ECoT) supervision to align cross-embodiment representations within the vision-language model (VLM). ZR-0 adopts a dual-stream architecture: a pre-trained VLM (System 2) generates structured ECoT reasoning during training, while a Diffusion Transformer-based action expert (System 1) produces continuous action chunks via flow matching. The two components are coupled through cross-attention, with an attention mask that restricts the action expert to input prompt features only, enabling ECoT generation to be entirely skipped at inference without any performance loss. ZR-0 is pre-trained on ProcCorpus-60M, a large-scale dataset comprising approximately 60 million frames (approximately 1,000 hours) from over 400K trajectories, with dense ECoT annotations covering 96.8% of all frames. We evaluate ZR-0 on three simulation benchmarks spanning single-arm (LIBERO), bimanual (RoboTwin 2.0), and humanoid (RoboCasa GR-1 Tabletop) embodiments, as well as real-world experiments on the xArm platform, demonstrating strong performance across all settings. Code and model checkpoints are available at https://github.com/RUCKBReasoning/ZR-0.

Related papers

• LM-X: Explainable Action Modeling with Progress, Event, and Uncertainty Prediction for Generalist Robot Manipulation [28.431160101245837]
  Generalist vision--language--action (VLA) policies learn long-horizon behavior mainly through short-horizon action prediction.<n>We introduce LM-X, which organizes prediction across task, event, and motor scales without claiming anatomical correspondence.
  arXiv  Detail & Related papers  (2026-08-26T13:05:29Z)
• G0.5: One Autoregressive Stream for Robot Reasoning and Action [43.43377463558725]
  We introduce G0.5, a pretrained autoregressive VLA in which a single transformer decoder emits reasoning and action tokens under a single objective.<n>Because reasoning and action share a single set of weights, the pretrained VLM's capabilities carry over to physical behavior.<n>G0.5 surpasses state-of-the-art models across 7 independent regimes.
  arXiv  Detail & Related papers  (2026-08-12T07:26:47Z)
• StellaVLA: In-Context Structured Demonstration for Generalizable Vision-Language-Action Models [37.45050514108603]
  We present StellaVLA, a framework that adapts at test time by conditioning on a single retrieved demonstration.<n>An automated offline pipeline converts each raw trajectory into a structured demonstration, e.g., a task plan, sub-goal descriptions, and verbalized 3D motion, at zero human-annotation cost.<n>A parallel dual-training design internalizes this reasoning during training through a joint action-and-language objective, while inference uses the action expert alone.
  arXiv  Detail & Related papers  (2026-08-12T05:30:53Z)
• GEAR-VLA: Learning Geometry-Aware Action Representations for Generalizable Robotic Manipulation [49.16739604572808]
  Vision-Language-Action (VLA) models achieve strong benchmark performance but struggle in real-world deployment with unseen objects.<n>We argue that this stems from the lack of a unified geometry-aware manipulation representation.<n>We propose GEAR-VLA, a VLA framework for learning unified geometry-aware action representations for generalizable robotic manipulation.
  arXiv  Detail & Related papers  (2026-06-07T09:23:16Z)
• RotVLA: Rotational Latent Action for Vision-Language-Action Model [54.22746299071677]
  We introduce RotVLA, a VLA framework built on a continuous rotational latent action representation.<n>Latent actions are modeled as elements of SO(n), providing continuity, compositionality, and structured geometry aligned with real-world action dynamics.<n>RotVLA consists of a VLM backbone and a flow-matching action head, pretrained on large-scale cross-embodiment robotic datasets and human videos.
  arXiv  Detail & Related papers  (2026-05-13T11:58:02Z)
• RLDX-1 Technical Report [74.70437517338186]
  We introduce RLDX-1, a general-purpose robotic policy for dexterous manipulation built on the Multi-Stream Action Transformer (MSAT)<n>Through empirical evaluation, we show that RLDX-1 consistently outperforms recent frontier Vision-Language-Action models.
  arXiv  Detail & Related papers  (2026-05-05T01:40:15Z)
• Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models [2.5338045371474816]
  Vision-Language-Action (VLA) models combine perception, language, and motor control in a single architecture.<n>We apply activation injection, sparse autoencoders, and linear probes to six models spanning 80M--7B parameters across 394,000+ rollout episodes on four benchmarks.
  arXiv  Detail & Related papers  (2026-03-19T17:59:55Z)
• From Passive Observer to Active Critic: Reinforcement Learning Elicits Process Reasoning for Robotic Manipulation [32.41847293364159]
  PRIMO R1 is a 7B framework that transforms video MLLMs into active "Critics"<n>We leverage outcome-based Reinforcement Learning to incentivize explicit Chain-of-Thought generation for progress estimation.<n>We establish state-of-the-art performance on RoboFail benchmark with 67.0% accuracy, surpassing closed-source models like OpenAI o1 by 6.2%.
  arXiv  Detail & Related papers  (2026-03-16T17:53:28Z)
• Universal Pose Pretraining for Generalizable Vision-Language-Action Policies [83.39008378156647]
  Existing Vision-Language-Action (VLA) models often suffer from feature collapse and low training efficiency.<n>We propose Pose-VLA, a decoupled paradigm that separates VLA training into a pre-training phase for extracting universal 3D spatial priors.<n>Our framework follows a two-stage pre-training pipeline, establishing fundamental spatial grounding via poses followed by motion alignment.
  arXiv  Detail & Related papers  (2026-02-23T11:00:08Z)
• Detect Anything via Next Point Prediction [51.55967987350882]
  Rex- Omni is a 3B-scale MLLM that achieves state-of-the-art object perception performance.<n>On benchmarks like COCO and LVIS, Rex- Omni attains performance comparable to or exceeding regression-based models.
  arXiv  Detail & Related papers  (2025-10-14T17:59:54Z)
• DexVLG: Dexterous Vision-Language-Grasp Model at Scale [59.5613919093295]
  There is little research on functional grasping with large models for human-like dexterous hands.<n>We introduce DexVLG, a large Vision-Language-Grasp model for Dexterous grasp pose prediction aligned with language instructions.<n>We generate a dataset of 170 million dexterous grasp poses mapped to semantic parts across 174,000 objects in simulation, paired with detailed part-level captions.
  arXiv  Detail & Related papers  (2025-07-03T16:05:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.