# ReasonX: MLLM-Guided Intrinsic Image Decomposition

arXiv: https://arxiv.org/abs/2512.04222 · HTML: https://mlsanity.com/paper/2512.04222/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 21 of its 56 references that are in the map.
- First posted on arXiv 3 Dec 2025.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)

## Builds on (most cited first, 21 of 21)

- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling](https://mlsanity.com/paper/2412.05271/) (2024)
- [Depth Anything V2](https://mlsanity.com/paper/2406.09414/) (2024)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [GeoWizard: Unleashing the Diffusion Priors for 3D Geometry Estimation from a Single Image](https://mlsanity.com/paper/2403.12013/) (2024)
- [OmniGen2: Towards Instruction-Aligned Multimodal Generation](https://mlsanity.com/paper/2506.18871/) (2025)
- [DIODE: A Dense Indoor and Outdoor DEpth Dataset](https://mlsanity.com/paper/1908.00463/) (2019)
- [Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence](https://mlsanity.com/paper/2505.23747/) (2025)
- [Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction](https://mlsanity.com/paper/2409.18124/) (2024)
- [Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning](https://mlsanity.com/paper/2310.12921/) (2023)
- [Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models](https://mlsanity.com/paper/2505.17015/) (2025)
- [Multimodal RewardBench: Holistic Evaluation of Reward Models for Vision Language Models](https://mlsanity.com/paper/2502.14191/) (2025)
- [Code as Reward: Empowering Reinforcement Learning with VLMs](https://mlsanity.com/paper/2402.04764/) (2024)
- [Open3D-VQA: A Benchmark for Comprehensive Spatial Reasoning with Multimodal Large Language Model in Open Space](https://mlsanity.com/paper/2503.11094/) (2025)
- [Visual Jigsaw Post-Training Improves MLLMs](https://mlsanity.com/paper/2509.25190/) (2025)
- [Enhancing Advanced Visual Reasoning Ability of Large Language Models](https://mlsanity.com/paper/2409.13980/) (2024)
- [Think Before You Diffuse: Infusing Physical Rules into Video Diffusion](https://mlsanity.com/paper/2505.21653/) (2025)
- [Dual-Process Image Generation](https://mlsanity.com/paper/2506.01955/) (2025)
- [PRISM: A Unified Framework for Photorealistic Reconstruction and Intrinsic Scene Modeling](https://mlsanity.com/paper/2504.14219/) (2025)
- [GeoPQA: Bridging the Visual Perception Gap in MLLMs for Geometric Reasoning](https://mlsanity.com/paper/2509.17437/) (2025)

## Built on by (most cited first, 5 of 5)

- [Advances in GRPO for Generation Models: A Survey](https://mlsanity.com/paper/2603.06623/) (2026)
- [MVID: Feed-Forward Multi-View Intrinsic Image Decomposition](https://mlsanity.com/paper/2512.23667/) (2025)
- [Geo-ID: Test-Time Geometric Consensus for Cross-View Consistent Intrinsics](https://mlsanity.com/paper/2603.13859/) (2026)
- [Albedo Estimation via Latent Bridge Matching](https://mlsanity.com/paper/2609.09884/) (2026)
- [EvenSplat: Coupled 2D-3D Decomposition for Gaussian Splatting under Exposure and Illumination Variation](https://mlsanity.com/paper/2610.01876/) (2026)

## Abstract (arXiv)

Intrinsic image decomposition aims to separate images into physical components such as albedo, depth, normals, and illumination. While recent diffusion- and transformer-based models benefit from paired supervision from synthetic datasets, their generalization to diverse, real-world scenarios remains challenging. We propose ReasonX, a novel framework that leverages a multimodal large language model (MLLM) as a perceptual judge providing relative intrinsic comparisons, and uses these comparisons as GRPO rewards for fine-tuning intrinsic decomposition models on unlabeled, in-the-wild images. Unlike RL methods for generative models, our framework aligns conditional intrinsic predictors by rewarding agreement between the judge's relational assessments and analytically derived relations from the model's outputs. ReasonX is model-agnostic and can be applied to different intrinsic predictors. Across multiple base architectures and modalities, ReasonX yields significant improvements, including 9-25% WHDR reduction on IIW albedo and up to 46% depth accuracy gains on ETH3D, highlighting the promise of MLLM-guided comparative supervision to bridge low- and high-level vision reasoning.