Tinghe Zhang

Researcher · Developer · Explorer

What I cannot create, I do not understand. — Richard Feynman

Research
AI
Interests
CoT, Security for AI, AI Merging, CV, Robotics
Now
Interning @ THU College of AI · Seeking for M.S./Phd this year
Tinghe Zhang

Education

High School Diploma

Liaoning Experimental High School

Bachelor

Northeastern University, Software College, Computer Security

Projects & Competitions

National College Student Innovation and Entrepreneurship Training Program

Completed Excellently

Training and using deep learning voiceprint recognition detection of cold-rolled production line faults

National Software Innovation Contest for College Students

National 1st Prize

Training and using deep learning voiceprint recognition detection of cold-rolled production line faults

National College Student Information Security Competition

National 1st Prize

Blockchain-based iris recognition technology using LWE

“Challenge Cup” – AI Main Competition

National 1st Prize / Champion

Developing an AI Agent to autonomously build an attack-defense test platform

Research & Publications

Intern @ MIT

Published

Advancing Alzheimer's PET Synthesis from MRI with Enhanced Pathological Feature Preservation

PWT: Advancing Alzheimer's PET Synthesis from MRI with Enhanced Pathological Feature Preservation

Qiaoru Li, Yujie Ren, Hao Zhang, Wenqiang Ge, Jiacheng Huang, Jinglin Xie, Danjie Cheng, Tinghe Zhang, Yitong Huang, Naidan Xu, Yuliang Gai, Longfei Zhou

IEEE 7th International Conference on Universal Village, 2024 · @ MIT

Paper 1 Paper 1

Research @ NEU

Published

Secure and Efficient Distributed Matrix Multiplication Based on Polynomial Coding

Secure and Efficient Distributed Matrix Multiplication Based on Polynomial Coding

Xiaowen Fan, Yunfeng Zhang, Tinghe Zhang, Qiang Wang

IEEE 7th International Conference on Next Generation Data-driven Networks, 2025 · @ Northeastern University

Paper 2 Paper 2

Collaborate with PKU

Published

Towards Zero-shot Cross-lingual SLU with Syntax-aware Multi-view Contrastive Learning

Towards Zero-shot Cross-lingual SLU with Syntax-aware Multi-view Contrastive Learning

Yuxin Xie, Zhen Xiong, Tinghe Zhang, Mengke Cui, Yuqi Li, Zhiqi Huang, Zhihong Zhu

IEEE International Conference on Acoustics, Speech and Signal Processing, 2025 · @ Peking University

Paper 3 Paper 3

Collaborate with SEU

Published

TOPOSHIELD: Reshaping the Flow of Malice via Spatio-Temporal Risk-Aware Topological Evolution in Multi-Agent Systems

TOPOSHIELD: Reshaping the Flow of Malice via Spatio-Temporal Risk-Aware Topological Evolution in Multi-Agent Systems

Ruiyang Huang, Chenxi Wang, Tinghe Zhang, Fengrui Liu, Jiayan Sun, Haocheng Wang, Yifan Wu

Published @ ACL Findings, 2026 · @ SEU

Paper 5 Paper 5

Reasearch @ NEU

Under Review

A Unified Framework for Trajectory Prediction with Explicit Planning and Reaction Decomposition

A Unified Framework for Trajectory Prediction with Explicit Planning and Reaction Decomposition

Jiaheng Chen, Jiaxing Li, Tinghe Zhang, Chao Peng Guo

Under review @ ACM MM , 2026 · @ Northeastern University

Paper 4 Paper 4

Intern @ THU College of AI

Ongoing

Reasoning Model&Embodied AI(Outside 3D Track in Autonomous Driving)

Scholarships

National Scholarship ¥10,000    0.2%

NEU First-Class Scholarship ¥2,000    3%

Activities

Served as :   Communist Youth League Branch Secretary   2024 Sep - Now

Served as :   Student Representative of College   2024 Nov - Now

Served as :   Vice Chair , 东北大学-华为奇点俱乐部   2025 May - Now

Forth , Software College “Freshman Cup” Basketball Tournament

Forth , Software College “Liu Changchun Cup” Basketball Tournament

Academic Pursuit

Reasoning in Large Language Models

Current

Focused on a Framework with SFT, RL, Test-Time Scaling to improve the reason ability of LLM

Embodied AI & Autonomous Driving

Current

Exploring the Autonomous Driving with the 3D Cloud as Input

Final Academic Pursuit (Which may seems to be too ambitious, but I Deeply believed in.)

Lifelong

1.Moving Beyond Statistical Correlation, Toward True "Understanding"​ :I am convinced that mainstream paradigms like the Transformer represent a highly efficient "statistical shortcut" rather than a genuine "path to comprehension." My core pursuit is to conceive a novel architecture that enables intrinsic learning and autonomous reasoning​ in machines—a formidable, perhaps even radical challenge, but one I believe is essential for achieving true machine intelligence.

2.Abandoning the "General" Mirage, Embracing "Domain" Breakthroughs:​ I do not pursue the AGI fantasy of "one model to solve everything." Instead, I find the vision of Specialist Superintelligence​ both feasible and compelling: building systems capable of assimilating knowledge, refining reasoning, and ultimately surpassing the boundaries of human expertise​ within a specific domain—and doing so in an extremely short timeframe. This is not about generality, but about profound, vertical evolution.

This journey will be long and fraught with fundamental challenges. That, however, is precisely what makes it meaningful. Perhaps one day, we will make it a reality.

Contact