DreamSAC: Learning Hamiltonian World Models via Symmetry Exploration

ArXi:2603.07545v1 Announce Type: cross 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