# InpaintDPO: Mitigating Spatial Relationship Hallucinations in Foreground-conditioned Inpainting via Diverse Preference Optimization

arXiv: https://arxiv.org/abs/2512.15644 · HTML: https://mlsanity.com/paper/2512.15644/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Diffusion and multimodal model alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 23 of its 75 references that are in the map.
- First posted on arXiv 16 Dec 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 23 of 23)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [SAM 2: Segment Anything in Images and Videos](https://mlsanity.com/paper/2408.00714/) (2024)
- [Decoupled Weight Decay Regularization](https://mlsanity.com/paper/1711.05101/) (2017)
- [IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models](https://mlsanity.com/paper/2308.06721/) (2023)
- [Qwen-Image Technical Report](https://mlsanity.com/paper/2508.02324/) (2025)
- [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://mlsanity.com/paper/2506.15742/) (2025)
- [Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis](https://mlsanity.com/paper/2306.09341/) (2023)
- [PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis](https://mlsanity.com/paper/2310.00426/) (2023)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [DanceGRPO: Unleashing GRPO on Visual Generation](https://mlsanity.com/paper/2505.07818/) (2025)
- [BrushNet: A Plug-and-Play Image Inpainting Model with Decomposed Dual-Branch Diffusion](https://mlsanity.com/paper/2403.06976/) (2024)
- [A Task is Worth One Word: Learning with Task Prompts for High-Quality Versatile Image Inpainting](https://mlsanity.com/paper/2312.03594/) (2023)
- [Bilateral Reference for High-Resolution Dichotomous Image Segmentation](https://mlsanity.com/paper/2401.03407/) (2024)
- [HD-Painter: High-Resolution and Prompt-Faithful Text-Guided Image Inpainting with Diffusion Models](https://mlsanity.com/paper/2312.14091/) (2023)
- [FreeEdit: Mask-free Reference-based Image Editing with Multi-modal Instruction](https://mlsanity.com/paper/2409.18071/) (2024)
- [Imagen 3](https://mlsanity.com/paper/2408.07009/) (2024)
- [HiFlow: Training-free High-Resolution Image Generation with Flow-Aligned Guidance](https://mlsanity.com/paper/2504.06232/) (2025)
- [VirtualModel: Generating Object-ID-retentive Human-object Interaction Image by Diffusion Model for E-commerce Marketing](https://mlsanity.com/paper/2405.09985/) (2024)
- [IAR2: Improving Autoregressive Visual Generation with Semantic-Detail Associated Token Prediction](https://mlsanity.com/paper/2510.06928/) (2025)

## Abstract (arXiv)

Foreground-conditioned inpainting, which aims at generating a harmonious background for a given foreground subject based on the text prompt, is an important subfield in controllable image generation. A common challenge in current methods, however, is the occurrence of Spatial Relationship Hallucinations between the foreground subject and the generated background, including inappropriate scale, positional relationships, and viewpoints. Critically, the subjective nature of spatial rationality makes it challenging to quantify, hindering the use of traditional reward-based RLHF methods. To address this issue, we propose InpaintDPO, the first Direct Preference Optimization (DPO) based framework dedicated to spatial rationality in foreground-conditioned inpainting, ensuring plausible spatial relationships between foreground and background elements. To resolve the gradient conflicts in standard DPO caused by identical foreground in win-lose pairs, we propose MaskDPO, which confines preference optimization exclusively to the background to enhance background spatial relationships, while retaining the inpainting loss in the foreground region for robust foreground preservation. To enhance coherence at the foreground-background boundary, we propose Conditional Asymmetric Preference Optimization, which samples pairs with differentiated cropping operations and applies global preference optimization to promote contextual awareness and enhance boundary coherence. Finally, based on the observation that winning samples share a commonality in plausible spatial relationships, we propose Shared Commonality Preference Optimization to enhance the model's understanding of spatial commonality across high-quality winning samples, further promoting shared spatial rationality.