Elamur Research⌕
☰
home / papers / 2026 / E-ViC: Reasoning Beyond Text via Embodied Visual Chain for Spatial Intelligence

E-ViC: Reasoning Beyond Text via Embodied Visual Chain for Spatial Intelligence

Paper · 2026 · Junbo Qi, Yi Zhang, Hanchu Ni et al.

Summary

Vision-Language Models used as the reasoning core of robots often fail at precise spatial tasks. The authors argue that text-based Chain-of-Thought is a poor medium for geometry because language compresses away the pixel-level detail that manipulation needs. E-ViC (Embodied Visual Chain) moves part of the reasoning into the image itself. The model can call visual operations, such as zooming or drawing points, boxes, trajectories and 3D boxes, then inspect the edited image before committing to an answer.

Training uses agentic reinforcement learning in an R1-Zero-like setup with GRPO. No supervised reasoning trajectories are provided; the model is rewarded only for task outcomes. A multi-task gated reward pushes it to actually use the tools, and repeated "check then answer" behaviour emerges from this signal.

The training set has 24.4K samples split into grounding and cognitive tasks, filtered for verifiability. On external embodied benchmarks, E-ViC beats strong VLM baselines by 10.1% on average. This suggests that tool-mediated visual verification is a promising route to spatial intelligence for embodied systems.

Key points

  • Training data: 24.4K curated samples covering grounding tasks (point, bounding box, trajectory, placement region) and cognitive tasks (spatial relations, depth/distance, numerical regression). Samples were selected through verifiability-based filtering.
  • Tool suite: zoom_in, draw_point, draw_box, draw_trajectory and draw_3d_bbox. The gated reward gives 0 for invalid outputs, 0.5 for valid answers and 1 when tools are used.
  • Visual tools vs. text-only reasoning: a text-only CoT RL variant with identical GRPO training was compared against E-ViC. E-ViC scores +8.3% higher on ERQA Action and +11.8% higher on ERQA Point, and also gains on Where2Place.
  • Ablations:
    • Without the gate, the model collapses to single-turn answers within 20 steps and trails by 18.1% on ERQA Action.
    • With only draw_point available, ERQA Action peaks at 56.9% and then drops to 38.9%, a sign of overfitting.
    • The full model's reasoning depth peaks at about 3.2 turns and settles near 2.7 turns.
  • Limitations: multi-turn inference adds computational cost. Real-world robustness, for example to sensor noise and partial observability, has not been tested.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Peking UniversityBeijing · China244 works · 39,359 citations🎓Imperial College LondonLondon · United Kingdom48 works · 12,181 citations🎓Waseda UniversityTokyo · Japan20 works · 584 citations🎓Southwestern University of Finance and EconomicsChengdu · China4 works · 1 citation

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Junbo Qi, Yi Zhang, Hanchu Ni, Che Liu, Zhimin Yao, Ruilin Yang, Xiancong Ren, Liangjian Wen, Wei Ge, Yuya Ieiri, Osamu Yoshie, Yong Dai and Xiaozhu Ju. 2026. E-ViC: Reasoning Beyond Text via Embodied Visual Chain for Spatial Intelligence. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1870

@inproceedings{qi2026e,
  title = {{E-ViC: Reasoning Beyond Text via Embodied Visual Chain for Spatial Intelligence}},
  author = {Junbo Qi, Yi Zhang, Hanchu Ni, Che Liu, Zhimin Yao, Ruilin Yang, Xiancong Ren, Liangjian Wen, Wei Ge, Yuya Ieiri, Osamu Yoshie, Yong Dai and Xiaozhu Ju},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1870},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1870}
}

References

34 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, and 8 others. 2025. Qwen2.5-vl technical report. Preprint, arXiv:2502.13…
  2. Tri Dao, Daniel Y. Fu, Stefano Ermon, Atri Rudra, and Christopher Ré. 2022. Flashattention: Fast and memory-efficient exact attention with io-awareness. Preprint, arXiv:2205.14135. find
  3. David Eigen, Christian Puhrsch, and Rob Fergus. 2014. Depth map prediction from a single image using a multi-scale deep network. Advances in neural information processing systems, 27. find
  4. The pascal visual object classes (voc) challenge. International journal of computer vision, 88(2):303– 338. find
  5. Yue Fan, Xuehai He, Diji Yang, Kaizhi Zheng, Ching-Chen Kuo, Yuting Zheng, Sravana Jyothi Naraya-naraju, Xinze Guan, and Xin Eric Wang. 2025. Grit: Teaching mllms to think with images. arXiv preprint arXiv:2505.15879. — [arXiv:2505.15879](https://arxiv.org/abs/2505.15879) find
  6. Yushi Hu, Weijia Shi, Xingyu Fu, Dan Roth, Mari Ostendorf, Luke Zettlemoyer, Noah A Smith, and Ranjay Krishna. 2024a. Visual sketchpad: Sketching as a visual chain of thought for multimodal language models. Advances in Neural Information Processing Systems, 37:139348–139379. find
  7. Yushi Hu, Otilia Stretcu, Chun-Ta Lu, Krishnamurthy Viswanathan, Kenji Hata, Enming Luo, Ranjay Krishna, and Ariel Fuxman. 2024b. Visual program distillation: Distilling tools and programmatic reasoning into vision-language models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat… find
  8. Chi-Pin Huang, Yueh-Hua Wu, Min-Hung Chen, Yu-Chiang Frank Wang, and Fu-En Yang. 2025. Thinkact: Vision-language-action reasoning via reinforced visual latent planning. arXiv preprint arXiv:2507.16815. — [arXiv:2507.16815](https://arxiv.org/abs/2507.16815) find
  9. Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, and 1 others. 2024. Openai o1 system card. arXiv preprint arXiv:2412.16720.
  10. Yuheng Ji, Huajie Tan, Jiayu Shi, Xiaoshuai Hao, Yuan Zhang, Hengyuan Zhang, Pengwei Wang, Mengdi Zhao, Yao Mu, Pengju An, and 1 others. 2025. Robo-brain: A unified brain model for robotic manipulation from abstract to concrete. In Proceedings of the Computer Vision and Pattern Recognition Conferenc… find
  11. NVIDIA, :, Alisson Azzolini, Junjie Bai, Hannah Brandon, Jiaxin Cao, Prithvijit Chattopadhyay, Huayu Chen, Jinju Chu, Yin Cui, Jenna Diamond, Yifan Ding, Liang Feng, Francesco Ferroni, Rama Govindaraju, Jinwei Gu, Siddharth Gururani, Imad El Hanafi, Zekun Hao, and 35 others. 2025. Cosmos-reason1: Fr… — [arXiv:2503.15558](https://arxiv.org/abs/2503.15558) find
  12. OpenAI, :, Aaron Hurst, Adam Lerer, Adam P. Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, Aleksander M adry, Alex Baker-Whitcomb, Alex Beutel, Alex Borzunov, Alex Carney, Alex Chow, Alex Kirillov, and 401 others. 2024. Gpt-4o system card.…
  13. OpenAI. 2025. Gpt-5 system card. find
  14. Mark Everingham, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. 2010. find
  15. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. DeepSeekMath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300.
  16. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2024. Hybridflow: A flexible and efficient rlhf framework. arXiv preprint arXiv: 2409.19256.
  17. Huajie Tan, Yuheng Ji, Xiaoshuai Hao, Xiansheng Chen, Pengwei Wang, Zhongyuan Wang, and Shanghang Zhang. 2025. Reason-rft: Reinforcement fine-tuning for visual reasoning of vision language models. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find
  18. Haozhe Wang, Alex Su, Weiming Ren, Fangzhen Lin, and Wenhu Chen. 2025a. Pixel reasoner: Incentivizing pixel-space reasoning with curiosity-driven reinforcement learning. arXiv preprint arXiv:2505.15966. — [arXiv:2505.15966](https://arxiv.org/abs/2505.15966) find
  19. Weiyun Wang, Zhangwei Gao, Lixin Gu, Hengjun Pu, Long Cui, Xingguang Wei, Zhaoyang Liu, Linglin Jing, Shenglong Ye, Jie Shao, and 1 others. 2025b. Internvl3.5: Advancing open-source multimodal models in versatility, reasoning, and efficiency. arXiv preprint arXiv:2508.18265.
  20. Junfei Wu, Jian Guan, Kaituo Feng, Qiang Liu, Shu Wu, Liang Wang, Wei Wu, and Tieniu Tan. 2025. Reinforcing spatial reasoning in vision-language models with interwoven thinking and visual drawing. Preprint, arXiv:2506.09965. — [arXiv:2506.09965](https://arxiv.org/abs/2506.09965) find
  21. Jihan Yang, Shusheng Yang, Anjali W Gupta, Rilyn Han, Li Fei-Fei, and Saining Xie. 2025. Thinking in space: How multimodal large language models see, remember, and recall spaces. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 10632–10643. find
  22. Wentao Yuan, Jiafei Duan, Valts Blukis, Wilbert Pumacay, Ranjay Krishna, Adithyavairavan Murali, Arsalan Mousavian, and Dieter Fox. 2024. Robopoint: A vision-language model for spatial affordance prediction for robotics. Preprint, arXiv:2406.10721. find
  23. Yifu Yuan, Haiqin Cui, Yibin Chen, Zibin Dong, Fei Ni, Longxin Kou, Jinyi Liu, Pengyi Li, Yan Zheng, and Jianye Hao. 2025a. From seeing to doing: Bridging reasoning and decision for robotic manipulation. Preprint, arXiv:2505.08548. — [arXiv:2505.08548](https://arxiv.org/abs/2505.08548) find
  24. Yifu Yuan, Haiqin Cui, Yaoting Huang, Yibin Chen, Fei Ni, Zibin Dong, Pengyi Li, Yan Zheng, and Jianye Hao. 2025b. Embodied-r1: Reinforced embodied reasoning for general robotic manipulation. Preprint, arXiv:2508.13998. — [arXiv:2508.13998](https://arxiv.org/abs/2508.13998) find
  25. Jiahui Zhang, Yurui Chen, Yanpeng Zhou, Yueming Xu, Ze Huang, Jilin Mei, Junhui Chen, Yu-Jie Yuan, Xinyue Cai, Guowei Huang, Xingyue Quan, Hang Xu, and Li Zhang. 2025a. From flatland to space: Teaching vision-language models to perceive and reason in 3d. Preprint, arXiv:2503.22976. — [arXiv:2503.22976](https://arxiv.org/abs/2503.22976) find
  26. Jingyi Zhang, Jiaxing Huang, Sheng Jin, and Shijian Lu. 2024. Vision-language models for vision tasks: A survey. IEEE transactions on pattern analysis and machine intelligence, 46(8):5625–5644. find
  27. Yi Zhang, Che Liu, Xiancong Ren, Hanchu Ni, Shuai Zhang, Zeyuan Ding, Jiayu Hu, Hanzhe Shan, Zhenwei Niu, Zhaoyang Liu, and 1 others. 2025b. Pelican-vl 1.0: A foundation brain model for embodied intelligence. arXiv preprint arXiv:2511.00108. find
  28. Yi Zhang, Che Liu, Xiancong Ren, Hanchu Ni, Yingji Zhang, Shuai Zhang, Zeyuan Ding, Jiayu Hu, Haozhe Shan, Junbo Qi, and 1 others. 2025c. Bridging vlms and embodied intelligence with deliberate practice policy optimization. arXiv preprint arXiv:2511.16602. — [arXiv:2511.16602](https://arxiv.org/abs/2511.16602) find
  29. Yi Zhang, Qiang Zhang, Xiaozhu Ju, Zhaoyang Liu, Jilei Mao, Jingkai Sun, Jintao Wu, Shixiong Gao, Shihan Cai, Zhiyuan Qin, Linkai Liang, Jiaxu Wang, Yiqun Duan, Jiahang Cao, Renjing Xu, and Jian Tang. 2025d. Embodiedvsr: Dynamic scene graph-guided chain-of-thought reasoning for visual spatial tasks.… — [arXiv:2503.11089](https://arxiv.org/abs/2503.11089) find
  30. Qingqing Zhao, Yao Lu, Moo Jin Kim, Zipeng Fu, Zhuoyang Zhang, Yecheng Wu, Zhaoshuo Li, Qianli Ma, Song Han, Chelsea Finn, and 1 others. 2025. Cot-vla: Visual chain-of-thought reasoning for vision-language-action models. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages… find
  31. Yanli Zhao, Andrew Gu, Rohan Varma, Liang Luo, Chien-Chin Huang, Min Xu, Less Wright, Hamid Shojanazeri, Myle Ott, Sam Shleifer, and 1 others. 2023. Pytorch fsdp: experiences on scaling fully sharded data parallel. arXiv preprint arXiv:2304.11277. find
  32. Ziwei Zheng, Michael Yang, Jack Hong, Chenxiao Zhao, Guohai Xu, Le Yang, Chao Shen, and Xing Yu. 2025. Deepeyes: Incentivizing" thinking with images" via reinforcement learning. arXiv preprint arXiv:2505.14362. — [arXiv:2505.14362](https://arxiv.org/abs/2505.14362) find
  33. Enshen Zhou, Jingkun An, Cheng Chi, Yi Han, Shanyu Rong, Chi Zhang, Pengwei Wang, Zhongyuan Wang, Tiejun Huang, Lu Sheng, and 1 others. 2025. Ro-borefer: Towards spatial referring with reasoning in vision-language models for robotics. arXiv preprint arXiv:2506.04308. — [arXiv:2506.04308](https://arxiv.org/abs/2506.04308) find
  34. A objects of interest: the black plastic mirror at (103, 242), the black glossy window at (423, 344), and the metallic silver bolt at (341, 349). Qualitative Case Studies To illustrate the visual chain reasoning process enabled by E-ViC, we present three representative case studies demonstrating how… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/qi-2026-e-vic-reasoning-beyond-text/