DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models

TL;DR

DICE introduces a novel method for precise inversion and controllable editing in discrete diffusion models, enhancing data reconstruction and manipulation in image and text domains without predefined masks.

Contribution

DICE is the first approach to enable accurate inversion and flexible editing in discrete diffusion models like multinomial diffusion and masked language models.

Findings

Preserves high data fidelity during editing

Enables accurate reconstruction without predefined masks

Improves controllability in discrete data editing

Abstract

Discrete diffusion models have achieved success in tasks like image generation and masked language modeling but face limitations in controlled content editing. We introduce DICE (Discrete Inversion for Controllable Editing), the first approach to enable precise inversion for discrete diffusion models, including multinomial diffusion and masked generative models. By recording noise sequences and masking patterns during the reverse diffusion process, DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. We demonstrate the effectiveness of DICE across both image and text domains, evaluating it on models such as VQ-Diffusion, Paella, and RoBERTa. Our results show that DICE preserves high data fidelity while enhancing editing capabilities, offering new opportunities for fine-grained content manipulation in discrete spaces.