DreamSAC：通过对称性探索学习哈密顿世界模型
DreamSAC: Learning Hamiltonian World Models via Symmetry Exploration
用自监督对比目标从视角相关的原始像素中提取不变物理状态，可作为视觉表征学习的一种思路：把守恒律/对称性作为监督信号，提升模型对未见物理属性的外推能力。
原文依据 using a novel self-supervised contrastive objective to identify the invariant physical state from raw, view-dependent pixel observations
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论学习到的世界模型擅长内插泛化，但在向新物理属性进行外推泛化时表现不佳。这一局限源于它们学习的是统计相关性，而非环境底层的生成规则，例如物理不变性与守恒律。我们认为，学习这些不变性是实现稳健外推的关键。为此，我们首先提出对称性探索，这是一种无监督探索策略：智能体由基于哈密顿量的好奇心奖励所内在驱动，主动探测并挑战自身对守恒律的理解，从而收集具有物理信息量的数据。其次，我们设计了一个基于哈密顿量的世界模型，它从所收集的数据中学习，并采用一种新颖的自监督对比目标，从视角相关的原始像素观测中识别不变的物理状态。我们的框架 DreamSAC 在这种主动筛选的数据上训练，在需要外推的 3D 物理仿真任务上显著优于最优基线。
英文原文摘要 1078 字符
Learned world models excel at interpolative generalization but fail at extrapolative generalization to novel physical properties. This limitation arises because they learn statistical correlations rather than the environment's underlying generative rules, such as physical invariances and conservation laws. We argue that learning these invariances is key to robust extrapolation. To achieve this, we first introduce Symmetry Exploration, an unsupervised exploration strategy where an agent is intrinsically motivated by a Hamiltonian-based curiosity bonus to actively probe and challenge its understanding of conservation laws, thereby collecting physically informative data. Second, we design a Hamiltonian-based world model that learns from the collected data, using a novel self-supervised contrastive objective to identify the invariant physical state from raw, view-dependent pixel observations. Our framework, DreamSAC, trained on this actively curated data, significantly outperforms state-of-the-art baselines in 3D physics simulations on tasks requiring extrapolation.