Agent Banana: High-Fidelity Image Editing with Agentic Thinking and Tooling

2026-02-09

AI Summary

This work addresses key challenges in professional image editing—such as object distortion, over-editing from multi-turn interactions, and the disconnect between low-resolution evaluation protocols and real-world high-resolution workflows—by proposing a hierarchical agent-based planning-execution framework. The approach leverages structured memory and a Context Folding mechanism to compress interaction history for long-range control, while Image Layer Decomposition enables localized edits and native 4K output. Additionally, the authors introduce HDD-Bench, the first high-resolution conversational benchmark for image editing. Experimental results demonstrate that the proposed method significantly outperforms existing approaches on HDD-Bench, achieving state-of-the-art performance in multi-turn consistency (IC: 0.871) and background fidelity (SSIM-OM: 0.84, LPIPS-OM: 0.12), while remaining competitive in single-turn editing tasks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Multiagent Systems: Multiagent Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We study instruction-based image editing under professional workflows and identify three persistent challenges: (i) editors often over-edit, modifying content beyond the user's intent; (ii) existing models are largely single-turn, while multi-turn edits can alter object faithfulness; and (iii) evaluation at around 1K resolution is misaligned with real workflows that often operate on ultra high-definition images (e.g., 4K). We propose Agent Banana, a hierarchical agentic planner-executor framework for high-fidelity, object-aware, deliberative editing. Agent Banana introduces two key mechanisms: (1) Context Folding, which compresses long interaction histories into structured memory for stable long-horizon control; and (2) Image Layer Decomposition, which performs localized layer-based edits to preserve non-target regions while enabling native-resolution outputs. To support rigorous evaluation, we build HDD-Bench, a high-definition, dialogue-based benchmark featuring verifiable stepwise targets and native 4K images (11.8M pixels) for diagnosing long-horizon failures. On HDD-Bench, Agent Banana achieves the best multi-turn consistency and background fidelity (e.g., IC 0.871, SSIM-OM 0.84, LPIPS-OM 0.12) while remaining competitive on instruction following, and also attains strong performance on standard single-turn editing benchmarks. We hope this work advances reliable, professional-grade agentic image editing and its integration into real workflows.

Problem

• image editing
• multi-turn editing
• high-fidelity
• over-editing
• ultra high-definition

Innovation

• Context Folding
• Image Layer Decomposition
• Agentic Image Editing
• High-Fidelity Editing
• HDD-Bench