I am Ting Huang, a visiting researcher in Prof. Hao Tang's group at Peking University. I received my M.S. from Shanghai University of Engineering Science (SUES) and work on embodied AI.

My research focuses on spatial intelligence and generalist embodied systems, including 3D vision-language and vision-language-action (VLA) models for mobile robots. My long-term goal is to build systems that understand and interact with the physical world. Outside research, I enjoy sports and music.

Spatial Intelligence Embodied AI Generalist Robotics

🔥 News

2026.06🎉🎉 MobileVLA-R1 , ConsiSpace, and OpenGround were accepted to ECCV 2026.
2026.05🎉🎉 Our RL Position Paper was accepted to ICML 2026 Position Papers.
2026.02🎉🎉 Multigranularity-3DQA was accepted to Expert Systems with Applications.
2026.01💼💼 I started a research internship at SRI-Robot, working on embodied AI.
2026.01🎉🎉 We released 3D CoCa v2 , a generalizable 3D captioning framework.
2025.11🎉🎉 We released MobileVLA-R1 , an embodied model for mobile robots.
2025.11🎉🎉 3D CoCa was accepted to 3DV 2026.
2025.09🐧🐧 I joined Prof. Hao Tang's group at Peking University as a visiting researcher.
2025.09🎉🎉 We released Nav-R1 , an embodied foundation model.
2025.08🎉🎉 We released 3D-R1 , an open-source generalist model for unified 3D scene understanding.

📝 Publications

arXiv Preprint

3D CoCa v2: Contrastive Learners with Test-Time Search for Generalizable Spatial Intelligence

Hao Tang^‡★, Ting Huang^★, Zeyu Zhang^★

Paper Code

3D CoCa v2 extends 3D CoCa with an inference-only test-time search (TTS) module and an external LLM judge.

ECCV 2026

Selected Publication

MobileVLA-R1: Reinforcing Vision-Language-Action for Mobile Robots

Ting Huang^★, Dongjian Li^★, Rui Yang^★, Zeyu Zhang^★†, Zida Yang, Hao Tang^‡

Project Paper Code AI Era AK

MobileVLA-R1 bridges language-guided high-level reasoning and continuous low-level control for mobile robots.
Leveraging a large VLA chain-of-thought dataset and reinforcement learning to produce interpretable plans and robust real-world execution.

arXiv Preprint

Nav-R1: Reasoning and Navigation in Embodied Scenes

Qingxiang Liu^★, Ting Huang^★, Zeyu Zhang^★†, Hao Tang^‡

Project Paper Code AIxiv

Nav-R1 unifies dialogue, reasoning, planning, and navigation in a single embodied foundation model.
Using Nav-CoT-110K and GRPO-based RL rewards plus a Fast-in-Slow paradigm to achieve coherent long-horizon reasoning with low-latency control in 3D environments.

arXiv Preprint

3D-R1: Enhancing Reasoning in 3D VLMs for Unified Scene Understanding

Ting Huang^★, Zeyu Zhang^★†, Hao Tang^‡

Project Paper Code AIxiv AK

A pioneering 3D-VLM leverages reinforcement learning and dynamic view selection to enhance reasoning capabilities in 3D scene understanding.

3DV 2026

Selected Publication

3D CoCa: Contrastive Learners are 3D Captioners

Ting Huang^★, Zeyu Zhang^★†, Yemin Wang^★, Hao Tang^‡

Paper Code DOI

Proposes 3D CoCa, a unified framework that jointly performs contrastive 3D-text alignment and 3D caption generation within one architecture, instead of relying on a two-stage "proposal-then-caption" pipeline.

🥇 Awards & Scholarships

Dec. 2024 — Outstanding Master's Student Scholarship.
Dec. 2023 — Graduate Entrance Scholarship.
Oct. 2020 — First Place, ROBOCON National College Student Robot Competition.

📖 Education

Sep. 2023 – Jul. 2026 — M.S., Shanghai University of Engineering Science.
Sep. 2018 – Jun. 2022 — B.S., Hebei University of Engineering.

😊 Academic Services

Conference Reviewer: AAAI 2026, AAAI 2027, 3DV 2026, ICRA 2026