Cosmos-Reason1：从物理常识到具身推理
人工智能
2025-05-20 v3 计算机视觉与模式识别
机器学习
机器人学
摘要
物理AI系统需要在物理世界中感知、理解和执行复杂动作。本文提出了Cosmos-Reason1模型，能够通过长链条式推理过程理解物理世界并生成恰当的具身决策（例如下一步行动），以自然语言形式呈现。我们首先定义了物理AI推理的关键能力，侧重于物理常识和具身推理。为表征物理常识，我们使用分层本体结构捕获关于空间、时间和物理的基本知识。对于具身推理，我们采用二维本体结构，跨不同物理具象化进行泛化。基于这些能力，我们开发了两个多模态大型语言模型，分别为Cosmos-Reason1-7B和Cosmos-Reason1-56B。我们收集数据并在两个阶段进行训练：物理AI监督微调 (SFT) 和物理AI强化学习 (RL)。为评估我们的模型，我们根据我们的本体构建了针对物理常识和具身推理的全面基准测试。评估结果表明，物理AI SFT和RL均带来了显著的改进。为促进物理AI的发展，我们将代码和预训练模型在NVIDIA开放模型许可证下提供，地址为https://github.com/nvidia-cosmos/cosmos-reason1。
引用
@article{arxiv.2503.15558,
title = {Cosmos-Reason1: From Physical Common Sense To Embodied Reasoning},
author = {NVIDIA and : and Alisson Azzolini and Junjie Bai and Hannah Brandon and Jiaxin Cao and Prithvijit Chattopadhyay and Huayu Chen and Jinju Chu and Yin Cui and Jenna Diamond and Yifan Ding and Liang Feng and Francesco Ferroni and Rama Govindaraju and Jinwei Gu and Siddharth Gururani and Imad El Hanafi and Zekun Hao and Jacob Huffman and Jingyi Jin and Brendan Johnson and Rizwan Khan and George Kurian and Elena Lantz and Nayeon Lee and Zhaoshuo Li and Xuan Li and Maosheng Liao and Tsung-Yi Lin and Yen-Chen Lin and Ming-Yu Liu and Xiangyu Lu and Alice Luo and Andrew Mathau and Yun Ni and Lindsey Pavao and Wei Ping and David W. Romero and Misha Smelyanskiy and Shuran Song and Lyne Tchapmi and Andrew Z. Wang and Boxin Wang and Haoxiang Wang and Fangyin Wei and Jiashu Xu and Yao Xu and Dinghao Yang and Xiaodong Yang and Zhuolin Yang and Jingxu Zhang and Xiaohui Zeng and Zhe Zhang},
journal= {arXiv preprint arXiv:2503.15558},
year = {2025}
}