Saved in:
Bibliographic Details
Main Authors: Pachebat, Jean, Conforti, Giovanni, Durmus, Alain, Janati, Yazid
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.03234
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917121143865344
author Pachebat, Jean
Conforti, Giovanni
Durmus, Alain
Janati, Yazid
author_facet Pachebat, Jean
Conforti, Giovanni
Durmus, Alain
Janati, Yazid
contents We introduce iterative tilting, a gradient-free method for fine-tuning diffusion models toward reward-tilted distributions. The method decomposes a large reward tilt $\exp(λr)$ into $N$ sequential smaller tilts, each admitting a tractable score update via first-order Taylor expansion. This requires only forward evaluations of the reward function and avoids backpropagating through sampling chains. We validate on a two-dimensional Gaussian mixture with linear reward, where the exact tilted distribution is available in closed form.
format Preprint
id arxiv_https___arxiv_org_abs_2512_03234
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Iterative Tilting for Diffusion Fine-Tuning
Pachebat, Jean
Conforti, Giovanni
Durmus, Alain
Janati, Yazid
Machine Learning
We introduce iterative tilting, a gradient-free method for fine-tuning diffusion models toward reward-tilted distributions. The method decomposes a large reward tilt $\exp(λr)$ into $N$ sequential smaller tilts, each admitting a tractable score update via first-order Taylor expansion. This requires only forward evaluations of the reward function and avoids backpropagating through sampling chains. We validate on a two-dimensional Gaussian mixture with linear reward, where the exact tilted distribution is available in closed form.
title Iterative Tilting for Diffusion Fine-Tuning
topic Machine Learning
url https://arxiv.org/abs/2512.03234