PMosFM: Preconditioned Manifold Matching for One-Step Physics-Constrained Generation
PMosFM turns physics-constrained generation into a one-step transport-and-decode process.
The paper says its manifold decoder encodes the constraints directly, avoiding separate residual losses and terminal residual unrolling. It adds geometric and covariance preconditioning to improve conditioning while learning transport in intrinsic coordinates. At inference, the method uses one neural transport evaluation followed by physical decoding. The authors report lower training and sampling time than multi-step baselines at comparable physical and distributional fidelity. ArXiv · AI/CL/LG's note
The paper says its manifold decoder encodes the constraints directly, avoiding separate residual losses and terminal residual unrolling. It adds geometric and covariance preconditioning to improve conditioning while learning transport in intrinsic coordinates. At inference, the method uses one neural transport evaluation followed by physical decoding. The authors report lower training and sampling time than multi-step baselines at comparable physical and distributional fidelity. ArXiv · AI/CL/LG's note
score 4