Making Implicit Preservation Intent Explicit in Conversational Image Editing

2026-07-08

AI Summary

This work addresses the challenge in conversational image editing where occluded content not explicitly modified often fails to be accurately recovered due to ambiguous retention intent. To resolve this, the authors propose ReSpec, a training-free framework that explicitly captures implicit retention intentions by leveraging historical visual references and generating recovery-aware instructions. ReSpec integrates historical image state selection, in-context editing models, and structured instruction generation, uniquely anchoring content preservation to the editing history rather than the current image state. The authors introduce OCCUR-Bench, a new benchmark for evaluating occlusion-aware content retention, on which ReSpec demonstrates significantly improved fidelity in content recovery and temporal consistency, thereby validating the efficacy and superiority of history-based semantic preservation in interactive image editing.

Abstract

Conversational image editing requires preserving not only visible content, but also content that temporarily disappears across turns. When newly added or modified content occludes a previously visible region, that region should reappear if it was never semantically changed. However, existing systems often fail to recover such occluded-but-unchanged content, producing inconsistent or hallucinated results. We introduce OCCUR-Bench, a diagnostic benchmark for temporal preservation in conversational image editing. OCCUR-Bench provides diverse occlusion-and-revelation scenarios with historical restoration references, enabling evaluation of faithful restoration rather than plausible regeneration. We also propose ReSpec, a training-free framework that makes implicit preservation explicit by pairing restoration-aware instructions with historical visual references. Given an editing history, ReSpec identifies what should persist, selects the historical image state that provides missing visual evidence, and conditions an in-context editor on the resulting instruction and reference image. Experiments show that ReSpec improves restoration fidelity and temporal consistency on OCCUR-Bench, highlighting the need to ground preservation in editing history rather than only the current image.

Problem

• conversational image editing
• temporal preservation
• occlusion recovery
• content consistency
• editing history

Innovation

• conversational image editing
• temporal preservation
• occlusion recovery
• history-aware restoration
• training-free framework