PIRF: Physics-Informed Reward Fine-Tuning for Diffusion Models
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 t...
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Articolo |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2509.20570 http://arxiv.org/abs/2509.20570 |
| Tags: |
Nessun Tag, puoi essere il primo ad aggiungerne!!
|