SpatialFlow-GRPO: Where Spatial Credit Drives Image Editing

• URL: http://arxiv.org/abs/2606.26872v2
• Date: Fri, 26 Jun 2026 15:03:59 GMT
• Title: SpatialFlow-GRPO: Where Spatial Credit Drives Image Editing
• Abstract summary: We propose SpatialFlow-GRPO, a training framework that introduces spatially fine-grained reward feedback.<n>The framework converts region-aware rewards into semantic-region-level optimization signals and aligns region advantages with the corresponding latent positions during policy updates.<n>The results show that SpatialFlow-GRPO converts local feedback into spatially aligned update signals and improves editing quality.
• Score: 29.67956120826966
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent online reinforcement learning has substantially improved image editing quality. However, existing Flow-GRPO-style methods usually rely on a single whole-image reward, which makes fine-grained editing optimization difficult. We observe that a key obstacle in image editing is this spatial uniformity assumption: a whole-image reward cannot distinguish how different spatial regions contribute to image quality. To address this issue, we propose SpatialFlow-GRPO, a training framework that introduces spatially fine-grained reward feedback. The framework converts region-aware rewards into semantic-region-level optimization signals and aligns region advantages with the corresponding latent positions during policy updates. We also train a region-aware reward model, SFReward, construct SFReward-14K with region-annotated editing samples, and introduce MultiEditBench to evaluate multi-region editing ability. On OmniGen2 and FLUX.2-klein-4B, SpatialFlow-GRPO outperforms Flow-GRPO on GEdit-Bench, ImgEdit-Bench, and MultiEditBench. The results show that SpatialFlow-GRPO converts local feedback into spatially aligned update signals and improves editing quality.

Related papers

• SAM-Flow: Source-Anchored Masked Flow for Training-Free Image Editing [11.868477331606911]
  SAM-Flow is a source-anchored masked flow framework for localized training-free image editing.<n>It achieves accurate semantic editing while significantly improving background preservation.<n>It can be integrated with mainstream flow-matching backbones such as Stable Diffusion 3 and FLUX without any fine-tuning.
  arXiv  Detail & Related papers  (2026-06-04T14:36:19Z)
• MT-EditFlow: Reinforcement Learning for Multi-Turn Image Editing with Flow Matching [91.83651402045108]
  MT-EditFlow is a flow-matching reinforcement learning framework designed to optimize reward signals for sequential image editing.<n>We show that MT-EditFlow significantly improves performance across diverse base models.
  arXiv  Detail & Related papers  (2026-06-01T09:46:10Z)
• Edit-GRPO: A Locality-Preserving Policy Optimization Framework for Image Editing [95.29911956874389]
  We propose Edit-GRPO, preserving Locality while optimizing image editing.<n>By assigning region-specific optimization signals to edit and non-edit areas, Edit-GRPO aligns policy updates with the spatial structure of editing tasks.<n>This design effectively suppresses common artifacts such as context distortion and boundary inconsistency.
  arXiv  Detail & Related papers  (2026-05-16T12:05:39Z)
• Rethinking Where to Edit: Task-Aware Localization for Instruction-Based Image Editing [30.23368654238746]
  We propose a training-free, task-aware edit localization framework for instruction-based image editing.<n>Our framework consistently improves non-edit region consistency while maintaining strong instruction-following performance.
  arXiv  Detail & Related papers  (2026-04-22T07:08:01Z)
• Region-Constrained Group Relative Policy Optimization for Flow-Based Image Editing [2.096755686662369]
  We propose RC-GRPO-Editing, a region-constrained GRPO post-training framework for flow-based image editing under deterministic ODE sampling.<n>It suppresses background-induced nuisance variance to enable cleaner localized credit assignment, improving region instruction adherence and preserving non-target content.<n> Experiments on CompBench show consistent improvements in editing region instruction adherence and non-target preservation.
  arXiv  Detail & Related papers  (2026-04-10T14:58:15Z)
• CoCoEdit: Content-Consistent Image Editing via Region Regularized Reinforcement Learning [15.375069717719157]
  We present a post-training framework for Content-Consistent Editing (CoCoEdit)<n>We first augment existing editing datasets with refined instructions and masks, from which 40K diverse and high quality samples are curated as training set.<n>We then introduce a pixel-level similarity reward to complement MLLM-based rewards, enabling models to ensure both editing quality and content consistency during the editing process.
  arXiv  Detail & Related papers  (2026-02-15T09:36:54Z)
• Follow-Your-Shape: Shape-Aware Image Editing via Trajectory-Guided Region Control [52.87568958372421]
  Follow-Your-Shape is a training-free and mask-free framework that supports precise and controllable editing of object shapes.<n>We compute a Trajectory Divergence Map (TDM) by comparing token-wise velocity differences between the inversion and denoising paths.<n>Our method achieves superior editability and visual fidelity, particularly in tasks requiring large-scale shape replacement.
  arXiv  Detail & Related papers  (2025-08-11T16:10:00Z)
• Training-free Geometric Image Editing on Diffusion Models [53.38549950608886]
  We tackle the task of geometric image editing, where an object within an image is repositioned, reoriented, or reshaped.<n>We propose a decoupled pipeline that separates object transformation, source region inpainting, and target region refinement.<n>Both inpainting and refinement are implemented using a training-free diffusion approach, FreeFine.
  arXiv  Detail & Related papers  (2025-07-31T07:36:00Z)
• LIME: Localized Image Editing via Attention Regularization in Diffusion Models [69.33072075580483]
  This paper introduces LIME for localized image editing in diffusion models.<n>LIME does not require user-specified regions of interest (RoI) or additional text input, but rather employs features from pre-trained methods and a straightforward clustering method to obtain precise editing mask.<n>We propose a novel cross-attention regularization technique that penalizes unrelated cross-attention scores in the RoI during the denoising steps, ensuring localized edits.
  arXiv  Detail & Related papers  (2023-12-14T18:59:59Z)
• Region-Aware Diffusion for Zero-shot Text-driven Image Editing [78.58917623854079]
  We propose a novel region-aware diffusion model (RDM) for entity-level image editing. To strike a balance between image fidelity and inference speed, we design the intensive diffusion pipeline. The results show that RDM outperforms the previous approaches in terms of visual quality, overall harmonization, non-editing region content preservation, and text-image semantic consistency.
  arXiv  Detail & Related papers  (2023-02-23T06:20:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.