# Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow

arXiv: https://arxiv.org/abs/2610.20071 · HTML: https://mlsanity.com/paper/2610.20071/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 24 references that are in the map.
- First posted on arXiv 14 Oct 2026.

## Builds on (most cited first, 23 of 23)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Demystifying MMD GANs](https://mlsanity.com/paper/1801.01401/) (2018)
- [StarGAN v2: Diverse Image Synthesis for Multiple Domains](https://mlsanity.com/paper/1912.01865/) (2019)
- [Mean Flows for One-step Generative Modeling](https://mlsanity.com/paper/2505.13447/) (2025)
- [One Step Diffusion via Shortcut Models](https://mlsanity.com/paper/2410.12557/) (2024)
- [Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models](https://mlsanity.com/paper/2404.07724/) (2024)
- [Improved Mean Flows: On the Challenges of Fastforward Generative Models](https://mlsanity.com/paper/2512.02012/) (2025)
- [Align Your Steps: Optimizing Sampling Schedules in Diffusion Models](https://mlsanity.com/paper/2404.14507/) (2024)
- [Alleviating Exposure Bias in Diffusion Models through Sampling with Shifted Time Steps](https://mlsanity.com/paper/2305.15583/) (2023)
- [Flow Map Language Models: One-step Language Modeling via Continuous Denoising](https://mlsanity.com/paper/2602.16813/) (2026)
- [Decoupled MeanFlow: Turning Flow Models into Flow Maps for Accelerated Sampling](https://mlsanity.com/paper/2510.24474/) (2025)
- [Towards More Accurate Diffusion Model Acceleration with A Timestep Tuner](https://mlsanity.com/paper/2310.09469/) (2023)
- [MeanCache: From Instantaneous to Average Velocity for Accelerating Flow Matching Inference](https://mlsanity.com/paper/2601.19961/) (2026)
- [Optimal Stepsize for Diffusion Sampling](https://mlsanity.com/paper/2503.21774/) (2025)
- [STORK: Faster Diffusion And Flow Matching Sampling By Resolving Both Stiffness And Structure-Dependence](https://mlsanity.com/paper/2505.24210/) (2025)
- [Understanding, Accelerating, and Improving MeanFlow Training](https://mlsanity.com/paper/2511.19065/) (2025)
- [IntMeanFlow: Few-step Speech Generation with Integral Velocity Distillation](https://mlsanity.com/paper/2510.07979/) (2025)
- [Flow Map Denoisers: Traversing the Distortion-Perception Plane for Inverse Problems](https://mlsanity.com/paper/2606.19802/) (2026)
- [Optimize Your Sampling: Tuned Diffusion Sampling with Bayesian Optimization](https://mlsanity.com/paper/2608.18040/) (2026)
- [Error as Signal: Stiffness-Aware Diffusion Sampling via Embedded Runge-Kutta Guidance](https://mlsanity.com/paper/2603.03692/) (2026)
- [Three Creates All: You Only Sample 3 Steps](https://mlsanity.com/paper/2603.22375/) (2026)

## Abstract (arXiv)

MeanFlow predicts interval-average velocities, but conventional few-step sampling ties the interval queried from the model to the duration of the state update. We separate these two quantities at inference, fixing the model, update grid, and number of queries while varying the prediction horizon. Our central finding is that the horizon response depends on both query position and the states entering that query. Across seven independently trained CIFAR-10 checkpoints at classifier-free guidance (CFG) scale 2, a longer horizon worsens FID at early steps but improves it at the midpoint. To probe whether query time alone determines this response, we fix the midpoint query and current/subsequent guidance while changing only earlier CFG. In held-out factorial outcomes on five previously studied checkpoints, the mean midpoint horizon benefit increases from 0.109 to 1.308 FID, indicating sensitivity to the incoming-state distribution. A frozen four-query rule reduces mean CIFAR-10 FID by 2.023 without retraining or extra queries, including gains on five checkpoints trained after rule selection. A milder rule also improves mean FID by 1.041 on a stronger frozen grid, with fixed-rule FID gains on the evaluated public ImageNet-256 and AFHQ checkpoints. However, matched-budget calibration favors query-start shifting in historical CIFAR tests, and the tested Horizon allocation adds no FID benefit beyond Grid+Start on two public ImageNet models. Thus, prediction horizon is a distinct, position- and state-sensitive inference control, but not a universally preferred one.