Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulation
Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulation
This research explores embodied chain-of-thought (CoT) for robotic control, introducing ERVLA, a vision-language-action model that effectively integrates linguistic reasoning into robot manipulation. ERVLA uses a CoT-dropout strategy during training and predicts actions directly during inference, achieving 86.9% success on LIBERO-Plus and 53.2% on VLABench, with superior performance in out-of-distribution settings.
This paper revisits embodied chain-of-thought (CoT) for robot manipulation, emphasizing the need to ground high-level reasoning in concrete action guidance. The authors introduce ERVLA, a vision-language-action model that leverages CoT during training but bypasses autoregressive instability during inference, achieving state-of-the-art results on LIBERO-Plus and VLABench.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Robot trajectories
978,743 trajectories
Trajectory count
samples
226,300,000
data
2,592.5 hours
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2606.03784v3 Announce Type: replace Abstract: Embodied chain-of-thought (CoT) aims to bridge linguistic reasoning with robotic control, yet its effective form and integration remain underexplored. In this paper, we revisit embodied CoT for robotic control at an unprecedented scale. We curate the largest embodied CoT corpus to date, comprising 978,743 trajectories, 226.3M samples, and 2592.5 hours of data. Throu
Open source S2
ling with increasing pre-training data. ERVLA achieves state-of-the-art results on LIBERO-Plus with an 86.9\% success rate and reaches 53.2\% on VLABench, showcasing superior performance in out-of-distribution settings. Furthermore, ERVLA outperforms competitive state-of-the-art baselines in real-robot experiments, especially in handling semantic ambiguity and long-horizon tasks. Code, data, and model checkpoints are
Open source S5
available on the project website: https://taoshuaiz.github.io/ERVLA/.
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org