FUSE: FK-Steered Multi-Modal Flow Matching for Efficient Simulation-Based Posterior Estimation

2026-07-06

AI Summary

Existing simulation-based posterior estimation methods struggle to effectively model multimodal distributions and suffer from limited accuracy due to their neglect of structural discrepancies between parameters and observations. This work proposes the FUSE framework, which innovatively integrates a dual-path multimodal flow-matching architecture with a Feynman–Kac (FK)-guided sampling mechanism. The former preserves and dynamically fuses heterogeneous input features, while the latter leverages intermediate observation likelihoods to steer trajectory generation, jointly addressing the heterogeneity between parameter and observation spaces. Experiments demonstrate that FUSE outperforms state-of-the-art methods on standard simulation-based inference benchmarks, closely approximating ground-truth MCMC posteriors, and successfully resolves parameter degeneracies in exoplanet orbital estimation tasks that are intractable for existing approaches.

Abstract

Simulation-Based Inference (SBI) is critical for scientific discovery, with generative models offering a promising path toward efficient inference. However, existing methods struggle with effective multimodal modeling. They often rely on brute-force fusion strategies that ignore the structural disparities between parameters and observations, thus limiting estimation fidelity. In this work, we introduce FUSE (Feynman-Kac steered mUlti-modal flow matching for efficient Simulation-based posterior Estimation). Unlike prior work, FUSE employs a dual-track architecture that preserves the distinct features of multimodal inputs while facilitating dynamic interaction. Additionally, we propose an FK-steered sampling strategy that leverages intermediate observation likelihoods to guide the generative trajectories, effectively improving the sample quality during inference. Our approach outperforms state-of-the-art baselines on standard SBI benchmarks, producing posteriors that closely match ground-truth MCMC. Furthermore, in a real-world exoplanet orbital estimation task, FUSE successfully resolves complex parameter degeneracies that challenge existing methods, highlighting its potential to accelerate complex scientific discoveries in astrophysics and beyond.

Problem

• Simulation-Based Inference
• Multimodal Modeling
• Posterior Estimation
• Parameter-Observation Disparity

Innovation

• multi-modal flow matching
• Feynman-Kac steering
• simulation-based inference
• dual-track architecture
• posterior estimation