# Stable-Layers: Fine-Tuning Image Layer Decomposition Models with VLM-Scored Reinforcement Learning

arXiv: https://arxiv.org/abs/2605.30257 · HTML: https://mlsanity.com/paper/2605.30257/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 37 of its 39 references that are in the map.
- First posted on arXiv 28 May 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 37)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2304.05977/) (2023)
- [Training Diffusion Models with Reinforcement Learning](https://mlsanity.com/paper/2305.13301/) (2023)
- [Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis](https://mlsanity.com/paper/2306.09341/) (2023)
- [Diffusion Model Alignment Using Direct Preference Optimization](https://mlsanity.com/paper/2311.12908/) (2023)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models](https://mlsanity.com/paper/2305.16381/) (2023)
- [Directly Fine-Tuning Diffusion Models on Differentiable Rewards](https://mlsanity.com/paper/2309.17400/) (2023)
- [DanceGRPO: Unleashing GRPO on Visual Generation](https://mlsanity.com/paper/2505.07818/) (2025)
- [Aligning Text-to-Image Diffusion Models with Reward Backpropagation](https://mlsanity.com/paper/2310.03739/) (2023)
- [Rectified Flow: A Marginal Preserving Approach to Optimal Transport](https://mlsanity.com/paper/2209.14577/) (2022)
- [VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation](https://mlsanity.com/paper/2412.21059/) (2024)
- [Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward](https://mlsanity.com/paper/2404.01258/) (2024)
- [Transparent Image Layer Diffusion using Latent Transparency](https://mlsanity.com/paper/2402.17113/) (2024)
- [Self-Rewarding Vision-Language Model via Reasoning Decomposition](https://mlsanity.com/paper/2508.19652/) (2025)
- [CanvasVAE: Learning to Generate Vector Graphic Documents](https://mlsanity.com/paper/2108.01249/) (2021)
- [MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?](https://mlsanity.com/paper/2407.04842/) (2024)
- [GRPO-Guard: Mitigating Implicit Over-Optimization in Flow Matching via Regulated Clipping](https://mlsanity.com/paper/2510.22319/) (2025)
- [Generative Image Layer Decomposition with Visual Effects](https://mlsanity.com/paper/2411.17864/) (2024)
- [Qwen-Image-Layered: Towards Inherent Editability via Layer Decomposition](https://mlsanity.com/paper/2512.15603/) (2025)
- [TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for Robotics](https://mlsanity.com/paper/2602.19313/) (2026)
- [LayerDiff: Exploring Text-guided Multi-layered Composable Image Synthesis via Layer-Collaborative Diffusion Model](https://mlsanity.com/paper/2403.11929/) (2024)
- [LayerD: Decomposing Raster Graphic Designs into Layers](https://mlsanity.com/paper/2509.25134/) (2025)
- [LayerFusion: Harmonized Multi-Layer Text-to-Image Generation with Generative Priors](https://mlsanity.com/paper/2412.04460/) (2024)
- [DreamLayer: Simultaneous Multi-Layer Generation via Diffusion Mode](https://mlsanity.com/paper/2503.12838/) (2025)
- [LayeringDiff: Layered Image Synthesis via Generation, then Disassembly with Generative Knowledge](https://mlsanity.com/paper/2501.01197/) (2025)
- [PSDiffusion: Harmonized Multi-Layer Image Generation via Layout and Appearance Alignment](https://mlsanity.com/paper/2505.11468/) (2025)

## Built on by (most cited first, 1 of 1)

- [UniWorld-Design: From Pixel Generation to Layer-Native Design](https://mlsanity.com/paper/2608.03971/) (2026)

## Abstract (arXiv)

We present Stable-Layers, a reinforcement learning framework that eliminates the need for paired supervision by fine-tuning a pretrained layer decomposition model using only feedback from a vision-language model (VLM). Starting from Qwen-Image-Layered, we apply Flow-GRPO with LoRA adaptation, sampling multiple candidate decompositions per image, scoring them with a VLM, and optimising the policy from group-relative advantages. The key challenge lies in designing a reliable reward signal: VLMs scoring samples in isolation tend to compress their judgements into a narrow band, leaving GRPO with little within-group variance to learn from. We address this with a two-stage evaluation pipeline that pairs structured per-sample scoring across five edit-centric criteria with a grid-based calibration step in which the VLM re-scores all candidates side-by-side. Stable-Layers produces decompositions with stronger layer separation, fewer blank or artifact-heavy layers, and lower per-layer reconstruction error on the Crello dataset compared to the base model.