ParetoSlider: Diffusion Models Post-Training for Continuous Reward Control

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Golan, Shelly, Finkelson, Michael, Bereslavsky, Ariel, Nitzan, Yotam, Patashnik, Or
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911615360696320
author Golan, Shelly
Finkelson, Michael
Bereslavsky, Ariel
Nitzan, Yotam
Patashnik, Or
author_facet Golan, Shelly
Finkelson, Michael
Bereslavsky, Ariel
Nitzan, Yotam
Patashnik, Or
contents Reinforcement Learning (RL) post-training has become the standard for aligning generative models with human preferences, yet most methods rely on a single scalar reward. When multiple criteria matter, the prevailing practice of ``early scalarization'' collapses rewards into a fixed weighted sum. This commits the model to a single trade-off point at training time, providing no inference-time control over inherently conflicting goals -- such as prompt adherence versus source fidelity in image editing. We introduce ParetoSlider, a multi-objective RL (MORL) framework that trains a single diffusion model to approximate the entire Pareto front. By training the model with continuously varying preference weights as a conditioning signal, we enable users to navigate optimal trade-offs at inference time without retraining or maintaining multiple checkpoints. We evaluate ParetoSlider across three state-of-the-art flow-matching backbones: SD3.5, FluxKontext, and LTX-2. Our single preference-conditioned model matches or exceeds the performance of baselines trained separately for fixed reward trade-offs, while uniquely providing fine-grained control over competing generative goals.
format Preprint
id arxiv_https___arxiv_org_abs_2604_20816
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ParetoSlider: Diffusion Models Post-Training for Continuous Reward Control
Golan, Shelly
Finkelson, Michael
Bereslavsky, Ariel
Nitzan, Yotam
Patashnik, Or
Machine Learning
Computer Vision and Pattern Recognition
Reinforcement Learning (RL) post-training has become the standard for aligning generative models with human preferences, yet most methods rely on a single scalar reward. When multiple criteria matter, the prevailing practice of ``early scalarization'' collapses rewards into a fixed weighted sum. This commits the model to a single trade-off point at training time, providing no inference-time control over inherently conflicting goals -- such as prompt adherence versus source fidelity in image editing. We introduce ParetoSlider, a multi-objective RL (MORL) framework that trains a single diffusion model to approximate the entire Pareto front. By training the model with continuously varying preference weights as a conditioning signal, we enable users to navigate optimal trade-offs at inference time without retraining or maintaining multiple checkpoints. We evaluate ParetoSlider across three state-of-the-art flow-matching backbones: SD3.5, FluxKontext, and LTX-2. Our single preference-conditioned model matches or exceeds the performance of baselines trained separately for fixed reward trade-offs, while uniquely providing fine-grained control over competing generative goals.
title ParetoSlider: Diffusion Models Post-Training for Continuous Reward Control
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.20816