Enregistré dans:
Détails bibliographiques
Auteurs principaux: Bahng, Hyojin, Chan, Caroline, Durand, Fredo, Isola, Phillip
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2506.02095
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911243951931392
author Bahng, Hyojin
Chan, Caroline
Durand, Fredo
Isola, Phillip
author_facet Bahng, Hyojin
Chan, Caroline
Durand, Fredo
Isola, Phillip
contents Measuring alignment between language and vision is a fundamental challenge, especially as multimodal data becomes increasingly detailed and complex. Existing methods often rely on collecting human or AI preferences, which can be costly and time-intensive. We propose an alternative approach that leverages cycle consistency as a supervisory signal. Given an image and generated text, we map the text back to image space using a text-to-image model and compute the similarity between the original image and its reconstruction. Analogously, for text-to-image generation, we measure the textual similarity between an input caption and its reconstruction through the cycle. We use the cycle consistency score to rank candidates and construct a preference dataset of 866K comparison pairs. The reward model trained on our dataset, CycleReward, outperforms state-of-the-art alignment metrics on detailed captioning, with superior inference-time scalability when used as a verifier for Best-of-N sampling, while maintaining speed and differentiability. Furthermore, performing DPO and Diffusion DPO using our dataset enhances performance across a wide range of vision-language tasks and text-to-image generation. Our dataset, model, and code are publicly released at https://cyclereward.github.io.
format Preprint
id arxiv_https___arxiv_org_abs_2506_02095
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cycle Consistency as Reward: Learning Image-Text Alignment without Human Preferences
Bahng, Hyojin
Chan, Caroline
Durand, Fredo
Isola, Phillip
Computer Vision and Pattern Recognition
Machine Learning
Measuring alignment between language and vision is a fundamental challenge, especially as multimodal data becomes increasingly detailed and complex. Existing methods often rely on collecting human or AI preferences, which can be costly and time-intensive. We propose an alternative approach that leverages cycle consistency as a supervisory signal. Given an image and generated text, we map the text back to image space using a text-to-image model and compute the similarity between the original image and its reconstruction. Analogously, for text-to-image generation, we measure the textual similarity between an input caption and its reconstruction through the cycle. We use the cycle consistency score to rank candidates and construct a preference dataset of 866K comparison pairs. The reward model trained on our dataset, CycleReward, outperforms state-of-the-art alignment metrics on detailed captioning, with superior inference-time scalability when used as a verifier for Best-of-N sampling, while maintaining speed and differentiability. Furthermore, performing DPO and Diffusion DPO using our dataset enhances performance across a wide range of vision-language tasks and text-to-image generation. Our dataset, model, and code are publicly released at https://cyclereward.github.io.
title Cycle Consistency as Reward: Learning Image-Text Alignment without Human Preferences
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2506.02095