FlowLensing: Simulating Gravitational Lensing with Flow Matching

2025-10-09

AI Summary

Existing gravitational lensing simulations rely on computationally expensive ray-tracing, achieving high fidelity but prohibitively low speed—hindering large-scale, high-resolution modeling of dark matter substructure. Method: We propose the first Diffusion Transformer-based flow-matching model tailored for strong gravitational lens image generation—the first application of flow matching in this domain—unifying discrete (e.g., dark matter model classes) and continuous (e.g., mass, position) physical parameter modeling while enforcing physical consistency in generated images. Contribution/Results: Our end-to-end framework synthesizes high-fidelity lensed images at over 200× speedup versus traditional simulators under complex dark matter models, with low inference latency and strong scalability. This work breaks the simulation bottleneck impeding fine-grained dark matter structure detection in cosmological surveys.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Gravitational lensing is one of the most powerful probes of dark matter, yet creating high-fidelity lensed images at scale remains a bottleneck. Existing tools rely on ray-tracing or forward-modeling pipelines that, while precise, are prohibitively slow. We introduce FlowLensing, a Diffusion Transformer-based compact and efficient flow-matching model for strong gravitational lensing simulation. FlowLensing operates in both discrete and continuous regimes, handling classes such as different dark matter models as well as continuous model parameters ensuring physical consistency. By enabling scalable simulations, our model can advance dark matter studies, specifically for probing dark matter substructure in cosmological surveys. We find that our model achieves a speedup of over 200$ imes$ compared to classical simulators for intensive dark matter models, with high fidelity and low inference latency. FlowLensing enables rapid, scalable, and physically consistent image synthesis, offering a practical alternative to traditional forward-modeling pipelines.

Problem

• Simulating high-fidelity gravitational lensing images efficiently
• Overcoming slow ray-tracing and forward-modeling bottlenecks
• Enabling scalable dark matter substructure studies through rapid simulation

Innovation

• Flow matching model for lensing simulation
• Handles discrete and continuous dark matter regimes
• Achieves 200x speedup over classical simulators