Research map / Image generation / Text-to-image diffusion
SpatialReward: Verifiable Spatial Reward Modeling for Fine-Grained Spatial Consistency in Text-to-Image Generation
Sashuai Zhou, Qiang Zhou, Junpeng Ma, Yue Cao, Ruofan Hu, Ziang Zhang, Xiaoda Yang, Zhibin Wang, Jun Song, Cheng Yu, Bo Zheng, Zhou Zhao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use SpatialReward as RL reward | SD3.5-M with Flow-GRPO · GenEval (80-Obj) | overall score: 0.67 -> 0.95 | Improved |
| use SpatialReward as RL reward | SD3.5-M with Flow-GRPO · SpatRelBench (1k-Obj) | overall score: 0.23 -> 0.42 | Improved |
| use SpatialReward as RL reward | FLUX1-dev with Flow-GRPO · GenEval (80-Obj) | overall score: 0.76 -> 0.97 | Improved |
| use SpatialReward as RL reward | FLUX1-dev with Flow-GRPO · SpatRelBench (1k-Obj) | overall score: 0.28 -> 0.46 | Improved |
| use SpatialReward instead of UnifiedReward | SD3.5-M with Flow-GRPO · SpatRelBench (1k-Obj) | overall score: 0.33 -> 0.42 | Improved |
| use QwenVL as the RL reward | SD3.5-M with Flow-GRPO · GenEval (80-Obj) | overall score: 0.67 -> 0.61 | Hurt |
| add exclusion constraints (ablation) | SpatialReward · SpatRelBench | accuracy: 25.9 -> 37.1 | Improved |
| add expert detection (ablation) | SpatialReward · GenEval | accuracy: 70.3 -> 95.2 | Improved |
| add expert detection (ablation) | SpatialReward · SpatRelBench | accuracy: 21.6 -> 37.1 | Improved |
| add chain-of-thought reasoning (ablation) | SpatialReward · SpatRelBench | accuracy: 27.9 -> 37.1 | Improved |
| add chain-of-thought reasoning (ablation) | SpatialReward · GenEval | accuracy: 94.2 -> 95.2 | Improved |
| train SD3.5-M with SpatialReward | SD3.5-M · Aesthetic | aesthetic score: 5.39 -> 5.23 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- + ImageReward — ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation (NeurIPS 2023) · better on 2 of 2 shared results
- VisionReward — VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation (AAAI 2026) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SpatialReward: Bridging the Perception Gap in Online RL for Image Editing via Explicit Spatial Reasoning (ICML 2026 · similarity 0.90)
- GoT-R1: Unleashing Reasoning Capability of Autoregressive Visual Generation with Reinforcement Learning (ICLR 2026 · similarity 0.89)
- Exploring Spatial Intelligence from a Generative Perspective (CVPR 2026 · similarity 0.88)
- Spatial Transport Optimization by Repositioning Attention Map for Training-Free Text-to-Image Synthesis (CVPR 2025 · similarity 0.87)
- 3D Space as a Scratchpad for Editable Text-to-Image Generation (CVPR 2026 · similarity 0.87)
- Focus-N-Fix: Region-Aware Fine-Tuning for Text-to-Image Generation (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
