Think3D introduces a framework that endows Vision‑Language Models with interactive 3D chain‑of‑thought reasoning by integrating 3D manipulation tools for active spatial exploration. The approach improves performance on benchmarks such as BLINK Multi‑view, MindCube‑1K, and VSI‑Bench‑Tiny for proprietary models like GPT‑4.1 and Gemini 2.5 Pro, and a reinforcement‑learning variant, Think3D‑RL, enables open‑weight models such as Qwen3‑VL‑4B to autonomously learn effective 3D exploration strategies, yielding tool‑use patterns comparable to stronger models and turning a performance drop on MindCube‑1K into a substantial improvement.
By Zaibin Zhang, Yuhan Wu, Lianjie Jia, Yifan Wang, Zhongbo Zhang, Yijiang Li, Binghao Ran, Fuxi Zhang, Zhuohan Sun, Yizhuang Peng, Zhenfei Yin, Lijun Wang, Huchuan Lu
arXiv:2606. 24472v1 Announce Type: cross Abstract: Vision-language-action (VLA) models have made rapid progress in generalist robot manipulation by harnessing semantic knowledge from pretrained vision-language backbones, but their visual tokens remain grounded in 2D image coordinates rather than the calibrated geometry of the robot's cameras -- a mismatch especially pronounced in multi-camera setups, where views are coupled by known intrinsics and extrinsics yet processed as independent images.
By Yue Peng, Yongzhe Zhao, Artur Habuda, Khuyen Pham, Yanheng Zhu, Tran Nguyen Le, Fares Abu-Dakka, Li Guo
The paper introduces Geometry-Privileged Distillation (GPD), a method that incorporates depth, semantic, and bird’s-eye-view cues as compact text routed to a teacher model during on-policy self-distillation. By applying a privileged KL loss only to incorrect trajectories, GPD enhances spatial reasoning in vision‑language models while keeping the deployed model RGB‑only. Experiments on a 4B backbone show that GPD outperforms prior methods on multiple spatial reasoning benchmarks, with ablations confirming the benefits of 3D cues, question‑conditioned routing, and selective distillation.
By Hongxing Li, Yixin Li, Dingming Li, Zixuan Wang, Yuchen Yan, Wenqi Zhang, Weiming Lu, Yongliang Shen
GaussVLA is a Vision‑Language‑Action model that enhances spatial reasoning by converting flat 2D visual tokens into compact 3D Gaussian tokens using a Gaussian Spatial Tokenizer. It further employs a Depth‑Aware Chain‑of‑Thought module to perform structured, non‑autoregressive geometric reasoning conditioned on language and flow‑time. In both simulated and real‑world tests, GaussVLA achieves high spatial‑manipulation success rates—93.5% on LIBERO and 100% on the Spatial suite—while using only 200 M parameters, outperforming SpatialVLA by 19.7% relative success.
By Md Selim Sarowar, Md Tanvir Islam, Sungho Kim, Sangtae Ahn
arXiv:2610.33558v1 Announce Type: new
Abstract: Video spatial reasoning requires multimodal large language models (MLLMs) to understand how objects are arranged and related across viewpoints. While t...
By Xinglin Li, TingTing Long, Jingzhi Zhou, Jiajing Chen, Jian Yang, Jin Xie
SpaceAct is a new benchmark that tests multimodal large language models on 3D spatial and temporal reasoning through closed‑loop physical action and active perception. It uses interlocking bricks and mechanical puzzles to create 108 verified instances across 36 task families, requiring models to reconstruct 3D geometry, anticipate dynamics, and execute precise 6‑DoF poses while avoiding collisions. The accompanying SAPA framework provides diagnostic modes and active‑perception capabilities, revealing that even with a valid plan, current models struggle with 3D visual grounding and action refinement.
By Chenglong Shi, Zekai Yuan, Yue Song, Jon Pu, Will Wu
arXiv:2606.22694v2 Announce Type: replace
Abstract: Vision-Language Models (VLMs) remain unreliable when spatial reasoning requires composing relations whose meanings depend on frames of reference. E...
By Danial Kamali, Tanawan Premsri, Shreya Rajpal, Amir Zadeh, Chuan Li, Parisa Kordjamshidi
arXiv:2610.18694v1 Announce Type: cross
Abstract: Vision-language models (VLMs) struggle with metric and multi-view relationships that remain implicit in RGB observations. Geometry foundation models...
By Rong Li, Shijie Li, Jiangyong Huang, Lingdong Kong, Jiaming Zhou, Dongyue Lu, Zeying Gong, Baoxiong Jia, Junwei Liang, Siyuan Huang
The paper introduces DeSpatial, a framework that separates spatial understanding into perception and reasoning. It uses a shared 3D perception tool with joint segmentation–depth co-refinement and cross-view instance fusion to produce a unified, metric, gravity-aligned 3D scene. DeSpatial achieves state‑of‑the‑art results on VSI‑Bench, MindCube‑Tiny, ViewSpatial, and VSI‑SUPER, demonstrating that the main bottleneck in spatial reasoning is precise 3D perception.
By Yifei Liu, Xiaolong Zhou, Ziyang Gong, Yifan Zhou, Yuanyuan Gao, Yuning Gong, Yuchen Yang, Hongjie Zhang, Xue Yang, Xiao Sun, Zhihang Zhong
AnchorVLN is an open‑vocabulary vision‑language navigation system that separates semantic proposals from geometric metrics. It uses a VLM to generate semantics while a geometry module supplies reliable metric quantities such as range and bearing, all within a Model Context Protocol server. The system achieves 64.4% on instruction following and improves object‑reference accuracy, reducing median center error from 3.37 m to 2.48 m.
By Long Giang Vu, Chengkai Yao, Yuxin Liu, FNU Aryan, Rajath Chandrashekar Aralikatti
The paper introduces FactoSR, a factorized reinforcement learning framework designed to improve spatial reasoning in Vision‑Language Models by addressing a dimensional mismatch between 2D visual inputs and the 3D+temporal nature of the physical world. FactoSR decomposes the reasoning task into three orthogonal geometric sub‑objectives—planar correspondence (XY), depth consistency (Z), and temporal reversibility (T)—and optimizes these constraints within a unified policy learning mechanism. Experiments on multi‑view and video benchmarks show that this decomposition yields significant performance gains, achieving a 5.9% improvement on VSI‑Bench and 4.5% on All‑Angles‑Bench.
By Yijun Yang, Shenghe Zheng, Wenbo Li, Jianhui Liu, Haoze Sun, Yanbing Zhang, Jiaxiu Jiang, Lin Song, Haoyang Huang, Nan Duan, Lei Zhu
OmniAct3D is a framework that adapts Vision Foundation Model (VFM) detectors trained on perspective images to equirectangular projection (ERP) panoramic images for 3D detection. It introduces the ERP-Ray Geometry Adapter (ERGA-Ray) to model spherical rays, the Visual-Action Reasoning Chain (VARC) to ground hypotheses in panoramic evidence, and the Appearance-Guided Heading Expert (AGHE) to recover high‑resolution heading cues. The method achieves a 2.96 NDS improvement on Spheriverse and a 24.87 mAP gain on PanoMMOcc, while preserving most of the baseline’s performance across configurations.
By Runtong Wu, Fei Teng, Di Wen, Guoqiang Zhao, Kunyu Peng, Kailun Yang