Paper page - Follow the Mean: Reference-Guided Flow Matching

Source: https://huggingface.co/papers/2605.10302

Abstract

Flow matching enables controllable generation through example-based adaptation via conditional endpoint mean adjustment, offering training-free and parametric guidance methods for style and content control.

Existing approaches to controllable generation typically rely on fine-tuning, auxiliary networks, or test-time search. We show thatflow matchingadmits a different control interface: adaptation through examples. Fordeterministic interpolants, thevelocity fieldis solely governed by aconditional endpoint mean; shifting this mean shifts the flow itself. This yields a simple principle for controllable generation: steer a pretrained model by changing the reference set it follows. We instantiate this idea in two forms.Reference-Mean Guidanceis training-free: it computes a closed-form endpoint-mean correction from a reference bank and applies it to a frozenFLUX.2-klein(4B) model, enabling control of color, identity, style, and structure while keeping the prompt, seed, and weights fixed.Semi-Parametric Guidanceamortizes the same idea through an explicit mean anchor and learned residual refiner, matching unconditionalDiT-B/4quality onAFHQv2while allowing the reference set to be swapped at inference time. These results point to a broader direction: generative models that adapt through data, not parameter updates.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2605\.10302

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.10302 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.10302 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.10302 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.