PIRF: Physics-Informed Reward Fine-Tuning for Diffusion Models
This addresses the issue of physical inconsistency in generative models for scientific applications, representing an incremental improvement over prior methods by enhancing efficiency and stability.
The paper tackled the problem of diffusion models producing outputs that violate physical laws by proposing Physics-Informed Reward Fine-tuning (PIRF), which frames physics-informed generation as a sparse reward optimization problem and bypasses value approximations to achieve superior physical enforcement across five PDE benchmarks.
Diffusion models have demonstrated strong generative capabilities across scientific domains, but often produce outputs that violate physical laws. We propose a new perspective by framing physics-informed generation as a sparse reward optimization problem, where adherence to physical constraints is treated as a reward signal. This formulation unifies prior approaches under a reward-based paradigm and reveals a shared bottleneck: reliance on diffusion posterior sampling (DPS)-style value function approximations, which introduce non-negligible errors and lead to training instability and inference inefficiency. To overcome this, we introduce Physics-Informed Reward Fine-tuning (PIRF), a method that bypasses value approximation by computing trajectory-level rewards and backpropagating their gradients directly. However, a naive implementation suffers from low sample efficiency and compromised data fidelity. PIRF mitigates these issues through two key strategies: (1) a layer-wise truncated backpropagation method that leverages the spatiotemporally localized nature of physics-based rewards, and (2) a weight-based regularization scheme that improves efficiency over traditional distillation-based methods. Across five PDE benchmarks, PIRF consistently achieves superior physical enforcement under efficient sampling regimes, highlighting the potential of reward fine-tuning for advancing scientific generative modeling.