Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Robust Amortized Simulation-Based Inference via Learned Error Models

NeurIPS 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Variational Bayesian deep learning

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes robust amortized simulation-based inference that learns an error model to handle misspecified simulators when multiple unlabeled observations are available (RVNPE).

Problem

Amortized simulation-based Bayesian inference with neural density estimators degrades significantly when the simulator does not accurately reflect the true data-generating process (model misspecification).

Approach

Introduces robust variational neural posterior estimation (RVNPE), which learns an error model from multiple unlabeled observations to account for simulator misspecification within amortized inference.

Experimental setup

Builds on: neural posterior estimation

Closest papers by meaning

• Amortized Simulation-Based Inference in Generalized Bayes via Neural Posterior Estimation (ICML 2026 · similarity 0.93)
• Adversarial robustness of amortized Bayesian inference (ICML 2023 · similarity 0.90)
• Leveraging Self-Consistency for Data-Efficient Amortized Bayesian Inference (ICML 2024 · similarity 0.89)
• All-in-one simulation-based inference (ICML 2024 · similarity 0.89)
• Variational Inference with Coverage Guarantees in Simulation-Based Inference (ICML 2024 · similarity 0.88)
• Addressing Misspecification in Simulation-based Inference through Data-driven Calibration (ICML 2025 · similarity 0.88)