RoboTwin

The 'RoboTwin' dataset/benchmark contains diverse robot data used to evaluate the performance of Vision-Language-Action models in unifying perception, language, and control for task execution.

Papers using RoboTwin (7)

X-Tokenizer: A Multimodal Action Tokenizer for Vision-Language-Action Pretraining2026Towards Long-horizon Embodied Agents with Tool-Aligned Vision-Language-Action Models2026VLA-Pro: Cross-Task Procedural Memory Transfer for Vision-Language-Action Models2026ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Models2026HALO: A Unified Vision-Language-Action Model for Embodied Multimodal Chain-of-Thought Reasoning2026Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment2025Light Future: Multimodal Action Frame Prediction Via Instructpix2pix2025