Organizations: National Tsing Hua University · National Yang Ming Chiao Tung University
Abstract
Vision-Language-Action (VLA) models leverage large-scale vision-language pretraining for flexible robot manipulation, yet at test time they remain brittle along two axes: spatial generalization, when object positions differ from those seen during training, and task generalization, when a familiar scene is paired with a different language instruction than the one seen in training. A growing family of methods addresses this brittleness by endowing a policy with the spatial and task-aware information such as 2D pixel-coordinate for object localization and placement. However, we find that existing representation through language prompting or visual prompting does not address the limitations; in contrast, exploiting a 3D point-based representation and feeding it directly to the action head leads to substantial improvements-revealing that how the grounding signal is represented and injected into the VLA is the true game changer. Thus, we propose a lightweight, model-agnostic module that represents the grounding signal in 3D, computes its relative displacement to the gripper, and injects the resulting spatial embedding directly into the action head through adaptive layer normalization. The entire module is a two-layer MLP that requires no changes to the VLA backbone or pretraining pipeline. On LIBERO-PRO, our method improves the average success rate of GR00T-N1.6 from 31.2 to 77.5 points under task perturbation and from 28.1 to 60.2 points under position perturbation (gains of 46.3 and 32.1 points). Comparable gains are achieved for π0.5 as well, demonstrating that the mechanism is backbone-agnostic. Together, these results support our central finding: given adequate grounding lifted into 3D, injecting it directly into the action head is what unlocks both spatial and task generalization in VLAs-achievable with nothing more than a lightweight module on top of a pretrained backbone.
Recent Vision-Language-Action (VLA) methods improve generalization by aligning their representations with 3D scene geometry. However, these methods are fundamentally instruction-agnostic: the representations align the entire scene uniformly, neglecting the 3D geometry of the specific target object designated by the language instruction. This causes failures on fine-grained manipulation and target occlusion tasks, where success depends on accurate 3D understanding of the target object rather than the entire scene. To address this, we present Mind-VLA, an instruction-aware spatial representation alignment method for VLA models. Specifically, Mind-VLA first obtains the target object specified by the language instruction, then prepares its target-object tri-view and extracts the corresponding VAE and VGGT features. Finally, the latent representation of the VLA model is aligned with these features to enable instruction-aware 3D understanding. Mind-VLA reaches 93.9% on LIBERO and 4.47 on CALVIN with a compact 345M-parameter backbone. On real-robot tasks with target occlusion, Mind-VLA reaches 54% average success, outperforming the best-performing instruction-agnostic method in real-robot comparison by 32 percentage points. Code will be publicly available.
Vision-Language-Action (VLA) models have shown strong potential for general-purpose robotic manipulation by leveraging large pretrained vision-language backbones. However, most existing VLAs rely primarily on 2D visual representations, which limit their ability to reason about fine-grained geometry and spatial grounding - capabilities that are essential for precise and robust manipulation in 3D environments. In this paper, we propose PointACT, a dual-system 3D-aware VLA policy that integrates hierarchical 3D point cloud representations directly into the action decoding process. PointACT employs a multi-scale point-action interaction mechanism with efficient bottleneck window self-attention, enabling evolving action tokens to densely attend to both local geometric detail and global scene structure. We evaluate PointACT on the LIBERO and RLBench benchmarks and systematically compare it against monolithic and dual-system VLA baselines, including variants augmented with point cloud inputs. PointACT achieves consistent improvements across both benchmarks, increasing success rates by 10% on the challenging RLBench-10Tasks suite over state-of-the-art pretrained VLAs, with even larger gains when the vision-language backbone is frozen and the action expert is trained from scratch. Extensive ablation studies demonstrate that tightly coupling hierarchical 3D geometry with pretrained 2D semantic representations is critical for robust and spatially grounded robot control. Our results also highlight the promise of pretrained 3D representations for 3D-aware VLA policies.
Existing Vision-Language-Action (VLA) models often suffer from feature collapse and low training efficiency because they entangle high-level perception with sparse, embodiment-specific action supervision. Since these models typically rely on VLM backbones optimized for Visual Question Answering (VQA), they excel at semantic identification but often overlook subtle 3D state variations that dictate distinct action patterns. To resolve these misalignments, we propose Pose-VLA, a decoupled paradigm that separates VLA training into a pre-training phase for extracting universal 3D spatial priors in a unified camera-centric space, and a post-training phase for efficient embodiment alignment within robot-specific action space. By introducing discrete pose tokens as a universal representation, Pose-VLA seamlessly integrates spatial grounding from diverse 3D datasets with geometry-level trajectories from robotic demonstrations. Our framework follows a two-stage pre-training pipeline, establishing fundamental spatial grounding via poses followed by motion alignment through trajectory supervision. Extensive evaluations demonstrate that Pose-VLA achieves state-of-the-art results on RoboTwin 2.0 with a 79.5% average success rate and competitive performance on LIBERO at 96.0%. Real-world experiments further showcase robust generalization across diverse objects using only 100 demonstrations per task, validating the efficiency of our pre-training paradigm.