PosteriorBench: From Point Estimates to Posterior Matching in Evaluating Generative Inverse Solvers

📄 arXiv: 2609.20794v1 📥 PDF

作者: Jiachen Yao, Zi-Siang Hsu, Xi Deng, Aditi Gupta, Xin Ju, Sally M Benson, Gege Wen, Anima Anandkumar

分类: cs.LG, cs.CE

发布日期: 2026-09-17

备注: 32 pages, 10 figures, 21 tables; the code is available at https://github.com/neuraloperator/PosteriorBench


💡 一句话要点

提出PosteriorBench以评估生成逆解算器的后验匹配

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 生成模型 逆问题 后验评估 分布匹配 神经算子 不确定性量化 物理模型

📋 核心要点

  1. 现有方法主要关注单一重建的准确性,未能有效处理病态问题中的多解情况。
  2. 提出PosteriorBench基准,通过高保真参考后验和五个评估指标,全面评估生成逆解算器的分布准确性。
  3. 实验表明,当前解算器在分布匹配上存在显著差距,神经算子提升了分辨率鲁棒性。

📝 摘要(中文)

生成模型在解决科学逆问题中越来越受到重视,但现有评估主要关注单一重建的可行性,这对病态问题而言是不够的。为此,本文提出了PosteriorBench,一个用于评估生成逆解算器分布准确性的基准。该基准涵盖四个基于物理的逆问题,并通过高保真参考后验进行评估,结合五个后验评估指标,全面考量点位准确性、边际不确定性、分布一致性和全局频率保真度。实验结果显示,当前解算器在分布匹配上存在显著差距,而神经算子在分辨率鲁棒性方面有所提升。

🔬 方法详解

问题定义:本文旨在解决生成逆解算器在处理病态问题时,无法有效捕捉真实后验分布的问题。现有方法往往只关注单一重建,导致模式崩溃或不确定性过于自信等问题。

核心思路:提出PosteriorBench基准,通过构建高保真参考后验,评估解算器是否能够恢复完整的解集,而不仅仅是单一最佳样本。

技术框架:PosteriorBench评估四个物理逆问题,采用重采样和马尔可夫链蒙特卡洛等计算密集型方法构建参考后验,并结合五个后验评估指标,形成统一的分布匹配和不确定性量化流程。

关键创新:最重要的创新在于引入了五个后验评估指标,全面评估解算器的分布准确性,解决了现有方法只关注点估计的局限性。

关键设计:在评估过程中,采用了后验均值误差、后验标准差误差、最大均值差异、切片Wasserstein距离和径向平均功率谱误差等指标,确保对点位准确性、边际不确定性、分布一致性和全局频率的全面考量。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,当前解算器在分布匹配上存在显著差距,具体而言,神经算子在分辨率鲁棒性方面表现出色,且引入的指导权重和生成噪声对后验方差的校准起到了关键作用。

🎯 应用场景

该研究的潜在应用领域包括科学计算、工程设计和环境监测等,能够为解决复杂的逆问题提供更为准确的后验分布评估,提升模型的可靠性和实用性。未来,该基准或将推动生成模型在更多科学领域的应用与发展。

📄 摘要(原文)

Generative models are increasingly used to solve scientific inverse problems, but existing evaluations still focus primarily on whether a method can produce a single plausible reconstruction. This is insufficient for ill-posed problems, where multiple solutions may be consistent with the same sparse or noisy observations. In these settings, a method can achieve strong pointwise accuracy while still failing to capture the true posterior through mode collapse, overconfident uncertainty, or averaging incompatible solutions. We introduce PosteriorBench, a benchmark for evaluating the distributional accuracy of generative inverse solvers. PosteriorBench evaluates four physics-based inverse problems: Darcy flow inversion, Poisson source recovery, carbon capture and storage, and light transport material inference. For each task, we construct high-fidelity reference posteriors using computationally heavy but established procedures such as rejection sampling and Markov chain Monte Carlo, enabling direct assessment of whether solvers recover the full set of solutions rather than the single best sample. We pair these references with a five-metric posterior evaluation suite: posterior-mean error, posterior-standard-deviation error, maximum mean discrepancy, sliced Wasserstein distance, and radially averaged power-spectrum error. These metrics assess pointwise accuracy, marginal uncertainty, distributional alignment, and global frequency fidelity. The benchmark spans sparse sensing, low-resolution observations, nonlinear forward models, varying noise levels, and multimodal priors, with a unified pipeline for distribution matching and uncertainty quantification. Our experiments reveal substantial distribution-matching gaps across current solvers, while showing that neural operators improve resolution robustness, and guidance weights and generation noise are key to posterior-variance calibration.