InternData-A1：面向通用策略预训练的高保真合成数据
InternData-A1: Pioneering High-Fidelity Synthetic Data for Pre-training Generalist Policy
提供了一套全解耦、可组合的仿真数据生成流水线，可批量产出多本体、多技能、长程操作轨迹；对做视觉-语言-动作模型的人而言，这是验证「纯合成数据预训练能否替代真机数据」的可复用数据与评测设定。
原文依据 synthetic data alone can match the performance of the strongest \pi-dataset in pre-training a VLA model
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期研究探讨了真实数据与合成数据如何影响 VLA 模型的泛化能力。尽管 \pi 系列模型已展现出大规模真机预训练的显著有效性，但合成数据此前尚未在规模上展现出可比的能力。本文首次给出证据，表明仅使用合成数据即可在 VLA 模型预训练中达到最强 \pi 数据集的性能，揭示了大规模仿真的巨大价值。所得模型在若干挑战性任务上还表现出出乎意料的强零样本仿真到现实迁移能力。我们的合成数据集 InternData-A1 包含超过 63 万条轨迹、7,433 小时，覆盖 4 种本体、18 种技能、70 个任务和 227 个场景，涵盖刚体、铰接体、可变形物体与流体对象的操作。该数据集通过一条高度自主、完全解耦且可组合的仿真流水线生成，支持灵活的任务组装、长程技能组合以及异构本体，且几乎无需人工调参。我们采用与 \pi_0 相同的架构，完全基于 InternData-A1 预训练一个模型，发现其在 49 个仿真任务、5 个真实世界任务和 4 个长程灵巧任务上与官方 \pi_0 相当。我们将开源该数据集与生成流水线，以拓宽大规模机器人数据的获取途径，并降低具身智能研究中可扩展数据创建的门槛。
英文原文摘要 1432 字符
Recent work explores how real and synthetic data contribute to VLA model generalization. While the \pi-series model has shown the strong effectiveness of large-scale real-robot pre-training, synthetic data has not previously demonstrated comparable capability at scale.This paper provides the first evidence that synthetic data alone can match the performance of the strongest \pi-dataset in pre-training a VLA model, revealing the substantial value of large-scale simulation.The resulting model also exhibits surprisingly strong zero-shot sim-to-real transfer on several challenging tasks.Our synthetic dataset, InternData-A1, contains over 630k trajectories and 7,433 hours across 4 embodiments, 18 skills, 70 tasks, and 227 scenes, covering rigid, articulated, deformable, and fluid-object manipulation. It is generated through a highly autonomous, fully decoupled, and compositional simulation pipeline that enables flexible task assembly, long-horizon skill composition, and heterogeneous embodiments with minimal manual tuning.Using the same architecture as \pi_0, we pre-train a model entirely on InternData-A1 and find that it matches the official \pi_0 across 49 simulation tasks, 5 real-world tasks, and 4 long-horizon dexterous tasks.We will open-source both the dataset and the generation pipeline to broaden access to large-scale robotic data and to lower the barrier to scalable data creation for embodied AI research.