---
title: "ROMS-IMLE: A Minimalist Approach to Competitive Single-Step Generative Modelling"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.19332v1"
permalink: "https://feed.astralor.com/items/itm_5b9493515a999bafab0ad638"
published: "2026-07-21T17:51:38.000Z"
score: 79
channel: "ai"
tags: ["ROMS-IMLE","One-step Generation","Convolutional network","Minimalist generative model","Implicit Maximum Likelihood Estimation (IMLE)"]
---

# ROMS-IMLE: A Minimalist Approach to Competitive Single-Step Generative Modelling

## AI commentary

> This paper directly confronts the 'stacking complexity' trend in generative models, achieving competitive single-step performance with the simplest IMLE+CNN. It's a must-read for engineers seeking efficient deployment solutions and theorists looking to challenge mainstream narratives.

## At a glance

This paper challenges the necessity of gradual noise transformation in generative models and proposes ROMS-IMLE, a minimalist framework. Using only Implicit Maximum Likelihood Estimation (IMLE) as the training objective paired with a moderately sized convolutional network, it avoids iterative denoising, transformer architectures, and complex inference. It achieves an FID of 2.56 on ImageNet 256 in a single step while maintaining good precision and recall, demonstrating that simplicity can yield competitive quality and speed.

## Key points

- Proposes ROMS-IMLE, a single-step generative model that completely avoids iterative denoising.
- Uses only IMLE training objective and a moderately sized convolutional network, without transformers or variational inference.
- Achieves FID 2.56 on ImageNet 256 with single-step generation, along with good precision and recall.
- Challenges the common belief that multi-step noise transformation is essential for high-performance generative models.

## Full text

Generative models have undergone many generations of evolution, from VAEs/GANs to diffusion/flow matching. Along the way, the underlying techniques have become more complicated and various beliefs about what drives strong empirical performance have taken hold. Due to the success of diffusion models and flow matching, one of the more common beliefs is the importance of transforming the noise distribution to the data distribution gradually through many small transformations. We ask whether this is truly necessary, and take a minimalist approach to designing a competitive generative model. We start with the bare-bones essentials, namely just a training objective and a model. We purposefully make both simple. For the training objective, we choose Implicit Maximum Likelihood Estimation (IMLE), and eschew more complicated alternatives such as variational inference, adversarial training and numerical integration. For the model, we eschew transformers and instead choose a moderately sized convolutional network. Then we judiciously added elements that are truly essential, which surprisingly do not include iterative denoising. The result is a single-step parameter-efficient generative model that produces high quality samples at fast speed: it achieves an FID of 2.56 on ImageNet 256 and simultaneously attains good precision and recall.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_5b9493515a999bafab0ad638