CVPR Findings2026
Thinking with Blueprints: Assisting Vision–Language Models in Spatial Reasoning via Structured Object Representation
Weijian Ma, Shizhao Sun, Tianyu Yu, Ruiyu Wang, Tat-Seng Chua, Jiang Bian
Abstract
techniques: (1) blueprint-embedded reasoning traces for supervised fine-tuning to elicit basic reasoning skills; (2) blueprint-aware rewards in reinforcement learning to encourage the blueprint to include an appropriate number of objects and to align final answers with this causal reasoning; and (3) anti-shortcut data augmentation that applies targeted perturbations to images and questions, discouraging reliance on superficial visual or linguistic cues. Experiments show that our method consistently outperforms existing VLMs and specialized spatial reasoning models.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
5 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
Cited by top-tier papers5
- HSSBench: Benchmarking Humanities and Social Sciences Ability for Multimodal Large Language ModelsZhaolu Kang, Junhao Gong, Jiaxu Yan, Wanke Xia et al.ICLR 2026 · 24 citations
- From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal ReasoningRuilin Luo, Chufan Shi, Yizhen Zhang, Cheng Yang et al.ICLR 2026 · 10 citations
- Android Coach: Improve Online Agentic Training Efficiency with Single State Multiple ActionsGuo Gan, Yuxuan Ding, Cong Chen, Yuwei Ren et al.ACL 2026 · 6 citations
- Bad Seeing or Bad Thinking? Rewarding Perception for Multimodal ReasoningHaozhe WANG, Qixin Xu, Changpeng Wang, Taofeng Xue et al.ICML 2026 · 3 citations
- MARIS: Marine Open-Vocabulary Instance SegmentationBingyu Li, Feiyu Wang, Da Zhang, Zhiyuan Zhao et al.CVPR 2026
Builds on25
- Video-R1: Reinforcing Video Reasoning in MLLMsKaituo Feng, Kaixiong Gong, Bohao Li, Zonghao Guo et al.NeurIPS 2025 · 528 citations
- Ferret: Refer and Ground Anything Anywhere at Any GranularityHaoxuan You, Haotian Zhang, Zhe Gan, Xianzhi Du et al.ICLR 2024 · 515 citations
- SpatialRGPT: Grounded Spatial Reasoning in Vision-Language ModelsAn-Chieh Cheng, Hongxu Yin, Yang Fu, Qiushan Guo et al.NeurIPS 2024 · 412 citations
- DeepEyes: Incentivizing "Thinking with Images" via Reinforcement LearningZiwei Zheng, Michael Yang, Jack Hong, Chenxiao Zhao et al.ICLR 2026 · 321 citations
- Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial IntelligenceDiankun Wu, Fangfu Liu, Yi-Hsin Hung, Yueqi DuanNeurIPS 2025 · 245 citations
Related papers
- Improve Vision Language Model Chain-of-thought ReasoningRuohong Zhang, Bowen Zhang, Yanghao Li, Haotian Zhang et al.ACL 2025 · 135 citations
- Unveiling the Compositional Ability Gap in Vision-Language Reasoning ModelTianle Li, Jihai Zhang, Yongming Rao, Yu ChengNeurIPS 2025 · 17 citations
- VTool-R1: VLMs Learn to Think with Images via Reinforcement Learning on Multimodal Tool UseMingyuan Wu, Jingcheng Yang, Jize Jiang, Meitang Li et al.ICLR 2026 · 87 citations
- RewardMap: Tackling Sparse Rewards in Fine-grained Visual Reasoning via Multi-Stage Reinforcement LearningSicheng Feng, Kaiwen Tuo, Song Wang, Lingdong Kong et al.ICLR 2026 · 28 citations
- Sparkle: Mastering Basic Spatial Capabilities in Vision Language Models Elicits Generalization to Spatial ReasoningYihong Tang, Ao Qu, Zhaokai Wang, Dingyi Zhuang et al.EMNLP Findings 2025 · 1 citation
