Paper 2605.03269
RLDX-1 Technical Report
- Published
- May 2026
- Research lab
- Independent
- Citations
- 11
- GitHub
- 326 stars
01 In brief
Summary
RLDX-1 is a general-purpose Vision-Language-Action model (VLA) for dexterous manipulation, integrating versatile intelligence with functional capabilities: motion awareness, long-term memory, and physical sensing.
Built on the Multi-Stream Action Transformer (MSAT) and a Qwen3-VL 8B backbone, it processes heterogeneous modalities via dedicated streams with joint self-attention.
Training uses public datasets (OXE, DROID, etc.), in-house data from ALLEX and Franka Research 3 platforms, and synthetic data generated via video models with motion-consistency filtering.
A three-stage pipeline (pre-training, mid-training, post-training) specializes the model, with optional reinforcement learning.
Inference optimizations (static graph conversion, kernel fusion) achieve 43.7 ms latency on an RTX 5090.
Evaluations show RLDX-1 outperforms baselines (π0.5, GR00T N1.6) on simulation benchmarks (e.g., 97.8% on LIBERO) and real-world tasks, notably 86.8% average success on ALLEX humanoid tasks versus ~40% for baselines, and 91.7% on Object-in-Box Selection requiring long-term memory.
The model also excels in contact-rich tasks like Plug Insertion and Egg Pick-and-Place using tactile and torque feedback.
RLDX-1 demonstrates that integrating functional capabilities is crucial for real-world dexterous manipulation beyond versatility alone.
02 From the paper
Abstract
While Vision-Language-Action models (VLAs) have shown remarkable progress toward human-like generalist robotic policies through the versatile intelligence (i.e. broad scene understanding and language-conditioned generalization) inherited from pre-trained Vision-Language Models, they still struggle with complex real-world tasks requiring broader functional capabilities (e.g. motion awareness, long-term memory, and physical sensing). To address this, we introduce RLDX-1, a general-purpose robotic policy for dexterous manipulation built on the Multi-Stream Action Transformer (MSAT), an architecture that unifies these capabilities by integrating heterogeneous modalities through modality-specific streams with cross-modal joint self-attention. RLDX-1 further combines this architecture with system-level design choices, including data synthesis for rare manipulation scenarios, learning procedures specialized for human-like manipulation, and inference optimizations for real-time deployment. Through empirical evaluation, we show that RLDX-1 consistently outperforms recent frontier VLAs (e.g. $π_{0.5}$ and GR00T N1.6) across both simulation benchmarks and real-world tasks that require broad functional capabilities beyond general versatility. In particular, RLDX-1 shows superiority in ALLEX humanoid tasks by achieving success rates of 86.8% while $π_{0.5}$ and GR00T N1.6 achieve around 40%, highlighting the ability of RLDX-1 to control a high-DoF humanoid robot under diverse functional demands. Together, these results position RLDX-1 as a promising step toward reliable VLAs for complex, contact-rich, and dynamic real-world dexterous manipulation.