Hi, thank you for the great work on RoboInter. We have been studying the paper closely while trying to reproduce the Planner training pipeline, and we realized that the exact training data composition is not fully clear to us from the paper.
Could you share the full dataset composition used for Planner training with the actual dataset names? In particular, we are wondering which datasets correspond to each category in Figure 11 (RoboInter-Spatial, RoboInter-Temporal, General Grounding, General Understanding, Simulation Data, Embodied Grounding, and Embodied Understanding), and what exact datasets are included in the general VQA data mentioned in the Appendix.
Thank you in advance.
Hi, thank you for the great work on RoboInter. We have been studying the paper closely while trying to reproduce the Planner training pipeline, and we realized that the exact training data composition is not fully clear to us from the paper.
Could you share the full dataset composition used for Planner training with the actual dataset names? In particular, we are wondering which datasets correspond to each category in Figure 11 (RoboInter-Spatial, RoboInter-Temporal, General Grounding, General Understanding, Simulation Data, Embodied Grounding, and Embodied Understanding), and what exact datasets are included in the general VQA data mentioned in the Appendix.
Thank you in advance.