💡 The paper introduces RLDX-1, a general-purpose robotic policy for dexterous manipulation that addresses the limitations of existing vision-language-action models. These models have shown progress in human-like generalist robotic policies but struggle with complex real-world tasks that require broader functional capabilities such as motion awareness, memory-aware decision making, and physical sensing. To overcome this, RLDX-1 uses a Multi-Stream Action Transformer architecture that integrates heterogeneous modalities through modality-specific streams with cross-modal joint self-attention. This architecture is combined with system-level design choices including synthesizing training data for rare manipulation scenarios, learning procedures specialized for human-like manipulation, and inference optimizations for real-time deployment. The results show that RLDX-1 outperforms recent frontier vision-language-action models across both simulation benchmarks and real-world tasks, achieving success rates of 86.8 percent in ALLEX humanoid tasks compared to around 40 percent for other models. This positions RLDX-1 as a promising step toward reliable vision-language-action models for complex and dynamic real-world dexterous manipulation. The method and results demonstrate the ability of RLDX-1 to control a high-degree-of-freedom humanoid robot under diverse functional demands, highlighting its potential for complex real-world tasks.

💡 The paper proposes a Geometric Action Model for robot policy learning that leverages pretrained geometric foundation models to enable language-conditioned manipulation policies in 3D physical environments. The problem addressed is that current vision-language-action models and video world-action models operate primarily on 2D image frames or 2D-derived latent spaces, leaving implicit the 3D geometry required for contact-rich manipulation.

The proposed method, Geometric Action Model, repurposes a pretrained geometric foundation model as a shared substrate for perception, temporal prediction, and action decoding. It splits the model at an intermediate layer, using the shallow layers as an observation encoder and inserting a causal future predictor to forecast future latent tokens conditioned on language, proprioception, and action history. The predicted future tokens are then routed through the remaining model blocks for feature propagation and decoding, allowing a single backbone to produce both future geometry and actions.

The results show that the Geometric Action Model is more accurate, more robust, faster, and lighter than current foundation-model-scale baselines across a broad suite of simulation and real-robot manipulation benchmarks. This design equips the geometric foundation model with language-conditioned temporal world modeling through minimal architectural modification while preserving its rich geometric priors, making it a significant contribution to robot policy learning in 3D physical environments.