Qinlei Xie | Dalian University of Technology

About Me

I am Qinlei Xie (谢钦磊), an undergraduate at Dalian University of Technology, majoring in Big Data Management and Application, with an expected graduation in June 2028. I previously conducted research at Tsinghua University under the guidance of Zheng Lu.

I am interested in learning world representations that improve how robots act. My broader interests include the action capabilities of vision-language-action models (VLAs), joint world-action representations, and video world models for robot learning. I am also interested in incorporating tactile and force feedback into manipulation policies to capture contact information that vision alone cannot reliably observe.

Research Interests

Generalizable robot policies: action representations, continuous action generation, and policy learning across objects, scenes, and tasks.
World-action representations: integrating latent dynamics and world representations with action prediction for closed-loop control.
Video world models: action-conditioned video prediction and its use in policy learning, planning, and interaction data generation.

Scope	Focus
Long-term goal	Build generalizable robot manipulation policies by jointly learning world representations and actions, incorporating visual, tactile, and force feedback.
Current work	Study action-conditioned residual latent dynamics in visual JEPA world models: predict state changes rather than full future representations, and evaluate their effect on planning and multi-step rollouts.
Current exploration	Explore the capabilities and limitations of LLMs in robotics, and how to improve latency and real-time responsiveness, fine-grained manipulation, and dexterous hand control.

I welcome discussions, suggestions, and guidance on these questions.

News

[2026] Towards Spatial Supersensing in the Wild was accepted to ECCV 2026.
[Jun. 2026] Our preprint Token Predictors Are Not Planners is available on arXiv.
[2026] GameVerse was accepted to ICML 2026.

Publications

arXiv

Token Predictors Are Not Planners: Building Physically Grounded Causal Reasoners

Zheng Lu*, Mingqi Gao*, Qinlei Xie*, Wanqi Zhong, Hanwen Cui, Heng Cao, Zirui Song, Yifan Yang, Chong Luo, Bei Liu, Yiming Li

arXiv preprint, 2026.

PDF Code * Equal contribution

ECCV

Towards Spatial Supersensing in the Wild

Tianjun Gu, Tianyu Xin, Kuan Zhang, Bowen Yang, Kok-Chung Chua, Peize Li, Xinran Zhang, Yupeng Chen, Qiyue Zhao, Qinlei Xie, Jianhang Liu, Yucheng Lu, Yinan Han, Marco Pavone, Yiming Li

European Conference on Computer Vision (ECCV), 2026.

PDF Code Project Page Poster

ICML

GameVerse: Can Vision-Language Models Learn from Video-based Reflection?

Kuan Zhang*, Dongchen Liu*, Qiyue Zhao, Jinkun Hou, Xinran Zhang, Qinlei Xie, Miao Liu, Yiming Li

International Conference on Machine Learning (ICML), 2026.

PDF Code Project Page

Education

Dalian University of Technology
B.S. in Big Data Management and Application
September 2024 - June 2028 (expected)

Hobbies

Piano: I have studied classical piano since childhood and passed the Grade 8 examination.
Basketball: I play as a swingman and enjoy shooting and driving to the basket.