Spatial-OPSD, a label-free self-improvement framework that instead exploits spatial structure naturally available from perception and reconstruction tools, is introduced, achieving the highest average among the open models and the best results on three of five spatial reasoning benchmarks.
Abstract
Vision-language models (VLMs) increasingly operate in embodied and spatially grounded settings, where accurate understanding of depth, viewpoint, and three-dimensional relations is essential. However, improving spatial reasoning typically relies on ground-truth answers, answer-derived rewards, or other forms of task-specific supervision. We introduce Spatial-OPSD, a label-free self-improvement framework that instead exploits spatial structure naturally available from perception and reconstruction tools. During training, a privileged teacher receives automatically obtainable spatial priors, such as depth, reconstructed 3D relations, and camera geometry, while the student observes only the original visual-language input. On trajectories sampled by the student itself, the teacher provides dense token-level supervision, allowing the student to internalize spatial knowledge without ground-truth answer labels or privileged information at inference time. To extend this supervision beyond a single round, we adopt a round-wise recursive training scheme: the teacher remains frozen within each round to provide a stable learning target, and the improved student initializes both teacher and student in the next round, where privileged spatial priors re-establish an informative teacher--student asymmetry. This enables repeated self-improvement while avoiding a rapidly moving teacher during optimization. Across four VLM families, a single round of Spatial-OPSD consistently improves the five-benchmark average, while three rounds further push a strong spatially specialized model to the open-source frontier, achieving the highest average among the open models and the best results on three of five spatial reasoning benchmarks. Our code is available at https://github.com/vermouth599/Spatial-OPSD.
GPD (Geometry-Privileged Distillation) is proposed, which makes geometric evidence the privilege in on-policy self-distillation (OPSD), outperforming both GRPO and answer-privileged OPSD across spatial reasoning benchmarks.
Hong-Xing Li, Yi-Xin Li, Ding-Ming Li et al.· 0 citations
Results show that spatially grounded privileged information can induce broader perceptual capabilities through on-policy self-distillation, enabling substantial synthetic-to-real transfer beyond the task and data distribution used for post-training.
Sophia Sirko-Galouchenko, M. Wysoczańska, Andrei Bursuc et al.· 0 citations
TTL-SR is proposed, a geometry-aware Test-Time Learning framework for quantitative Spatial Reasoning that leverages geometric consistency constraints and unlabeled test data to adapt models to target domains to boost spatial reasoning performance.
Ge-Ge Zhang, Shuai-Cheng Niu, Gang Dai et al.· 0 citations
SIG is introduced, a structured grid-based schema that explicitly represents object layouts, inter-object relations, and physically grounded priors that demonstrates the generalizability of SIG across diverse scenarios and multiple MLLMs, and systematically examines how data sample selection influences SIG-based ICL.
Guan-Lin Wu, Bo-Yan Su, Yang Zhao et al.· IEEE Transactions on Pattern...· 0 citations
SpatialSpeak is introduced, a two-stage framework that connects QA-native reconstruction pretraining with spatial CoT learning and achieves state-of-the-art results on ReVSI, VSI-Bench, and SPAR-Bench, with a ReVSI score of 62.8 that exceeds the strongest compared baseline by 8.7 points.
Yang Cao, Jia-Xin Zhang, Dave Zhenyu Chen et al.· 0 citations
Large Vision-Language Models (LVLMs) have made remarkable progress across visual perception tasks, yet spatial reasoning remains a persistent weakness, especially for questions that require reasoning over visual space. Recent spatial-reasoning methods incorporate generated grounding, where models predict bounding boxes...
Rafi Ibn Sultan, Xiang-Yu Zhou, Mohammad O. S. Chowdhury et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.