Text-guided image editing often suffers from global drift, where changing a prompt to modify a specific attribute (e.g., changing a dog to a cat) inadvertently alters the background or layout of the entire image. The authors investigate how to achieve targeted local edits in Diffusion Transformers (DiTs) without requiring external masks, user-provided guidance, or model fine-tuning.
ATDEdit reformulates the diffusion sampling process as an asynchronous, token-indexed operation. Instead of updating all latent tokens synchronously under the new target prompt, the framework uses a two-stage process:
Sensitivity Estimation: It calculates a "Token-wise Conditional Surprisal" score, which measures how much each token's prediction changes when the condition is switched. This identifies which tokens are most relevant to the requested edit.
Asynchronous Updating: The framework applies target-conditioned updates only to the high-surprisal (editable) tokens. For the remaining "keep" tokens, it enforces background preservation by projecting their latent states back to the source values and supplying their key/value memory from the source branch during self-attention.
This approach provides a flexible, training-free way to perform high-fidelity image editing. By moving away from static, pre-computed masks and toward dynamic, model-driven introspection, ATDEdit allows the editing process to adapt to the evolving semantics of the reverse diffusion trajectory. Experiments on the PIE-Bench dataset demonstrate that this method achieves superior preservation of the original image structure while maintaining competitive semantic alignment with the target text.
A sensitivity metric that quantifies how much a specific latent token's predicted velocity changes when the conditioning prompt is switched.
Asynchronous Token Decoding
A sampling strategy where different latent tokens receive distinct update rules (e.g., target-conditioned updates vs. source-memory preservation) within the same diffusion step.
Hard Projection
A mathematical operation that resets the latent states of selected "keep" tokens to their original source values at each sampling step to prevent background drift.
Common-noise Coupling
A technique that uses the same stochastic noise realization for both source and target branches to reduce variance in the counterfactual comparison between conditions.
Diffusion Transformers (DiTs)
A generative model architecture that uses transformer blocks to process latent tokens, serving as the backbone for the editing framework.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.