InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences

2025-03-14

AI Summary

Current plug-and-play diffusion priors (PnP-DP) lack systematic evaluation of generalization and robustness in scientific inverse problems. To address this, we introduce InverseBench—the first physics-informed benchmark for diffusion priors, covering optical tomography, medical imaging, black-hole imaging, seismology, and fluid dynamics, each with domain-specific structural constraints and physical modeling challenges. We integrate 14 diffusion-driven PnP algorithms within a unified framework featuring differentiable forward modeling, multi-scale sampling, and physics-constrained reconstruction. Comprehensive empirical evaluation reveals the performance boundaries and failure modes of PnP-DP relative to domain-specific traditional methods. All code, datasets, and pre-trained models are publicly released, establishing a standardized, reproducible platform for evaluating interpretable and trustworthy AI in scientific computing.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Data transparency and provenance

Abstract

Plug-and-play diffusion priors (PnPDP) have emerged as a promising research direction for solving inverse problems. However, current studies primarily focus on natural image restoration, leaving the performance of these algorithms in scientific inverse problems largely unexplored. To address this gap, we introduce extsc{InverseBench}, a framework that evaluates diffusion models across five distinct scientific inverse problems. These problems present unique structural challenges that differ from existing benchmarks, arising from critical scientific applications such as optical tomography, medical imaging, black hole imaging, seismology, and fluid dynamics. With extsc{InverseBench}, we benchmark 14 inverse problem algorithms that use plug-and-play diffusion priors against strong, domain-specific baselines, offering valuable new insights into the strengths and weaknesses of existing algorithms. To facilitate further research and development, we open-source the codebase, along with datasets and pre-trained models, at https://devzhk.github.io/InverseBench/.

Problem

• Evaluates diffusion models for scientific inverse problems.
• Benchmarks 14 algorithms using plug-and-play diffusion priors.
• Addresses unique structural challenges in five scientific domains.

Innovation

• Introduces InverseBench for scientific inverse problems
• Benchmarks 14 algorithms using plug-and-play diffusion priors
• Open-sources codebase, datasets, and pre-trained models