ProgressVLA: Progress-Guided Diffusion Policy for Vision-Language Robotic Manipulation
This work addresses the lack of progress awareness in VLA models for long-horizon robotic manipulation tasks, offering a differentiable progress guidance method that improves task success and generalization.
ProgressVLA introduces a progress-aware vision-language-action model for robotic manipulation, achieving a low prediction residual of 0.07 in simulation and zero-shot generalization to real-world tasks, with substantial improvements in success rates on CALVIN and LIBERO benchmarks.
Most existing vision-language-action (VLA) models for robotic manipulation lack progress awareness, typically relying on hand-crafted heuristics for task termination. This limitation is particularly severe in long-horizon tasks involving cascaded sub-goals. In this work, we investigate the estimation and integration of task progress, proposing a novel model named {\textbf \vla}. Our technical contributions are twofold: (1) \emph{robust progress estimation}: We pre-train a progress estimator on large-scale, unsupervised video-text robotic datasets. This estimator achieves a low prediction residual (0.07 on a scale of $[0, 1]$) in simulation and demonstrates zero-shot generalization to unseen real-world samples, and (2) \emph{differentiable progress guidance}: We introduce an inverse dynamics world model that maps predicted action tokens into future latent visual states. These latents are then processed by the progress estimator; by applying a maximal progress regularization, we establish a differentiable pipeline that provides progress-piloted guidance to refine action tokens. Extensive experiments on the CALVIN and LIBERO benchmarks, alongside real-world robot deployment, consistently demonstrate substantial improvements in success rates and generalization over strong baselines.