Jingfeng Zhang

Trustworthy Machine Learning · Safe Embodied AI

Building trustworthy AI for high-stakes decisions

Jingfeng Zhang · Associate Professor & PhD Supervisor
Fudan University

I lead RISE-AI Lab at Fudan University, where we study Trustworthy Machine Learning & Safe Embodied Intelligence

ICLR OralNeurIPS SpotlightRIKEN Ohbu Award

Apply to RISE-AI Explore the research
Google Scholar ↗ ORCID ↗ GitHub ↗

College of Intelligent Robotics and Advanced Manufacturing · Institute of Trustworthy Embodied AI

Portrait of Jingfeng Zhang
Based in Shanghai · working globally
30+ Papers at leading AI venues ICML, NeurIPS, ICLR and IEEE TPAMI Research Distinctions Oral · Spotlight · Best ICLR 2021 · NeurIPS 2023 · AAAI 2025 RIKEN Ohbu Award Research recognition RIKEN AIP, FY2021 8 countries Global experiences China · U.K. · U.S.A. · Singapore · Japan · New Zealand · Saudi Arabia · U.A.E.

Research with the TrustAI mandate

Trustworthy AI, from learning theories to embodied intelligence.

My research asks a practical question: how can an intelligent system stay dependable when its data are imperfect, its environment shifts, or an adversary actively tries to break it? I develop learning principles, evaluations, and defenses that make failure visible—and safety actionable.

I am an Associate Professor and PhD Supervisor at Fudan University. My primary career path includes a permanent lectureship at the University of Auckland and research appointments at RIKEN AIP, following a bachelor's degree from Shandong University and a PhD from the National University of Singapore. Complementary international experience includes research roles with NYU Abu Dhabi and KAUST, internships at IBM Research, Singapore and RIKEN AIP, part-time teaching at the University of Tokyo and Ochanomizu University, and undergraduate study at UCLA and the University of Cambridge.

Formal appointments & education

2026-CurrentAssociate Professor & PhD SupervisorFudan University 2023—2026Lecturer (permanent) & PhD SupervisorUniversity of Auckland 2021—2023Postdoctoral Researcher → Research ScientistRIKEN AIP 2016—2020PhD in Computer ScienceNational University of Singapore 2012—2016BSc in Computer Science & TechnologyTaishan College, Shandong University

Exchange, visiting, internships & part-time roles

2023-CurrentVisiting ScientistRIKEN AIP 2022—2026Guest Lecturer (part-time)The University of Tokyo 2026Guest Lecturer (part-time)Ochanomizu University 2024-2026Research AssociateCenter for Artificial Intelligence and Robotics (CAIR), New York University Abu Dhabi 2024-2026Visiting ProfessorPRADA Lab, King Abdullah University of Science and Technology (KAUST) 2019Research InternRIKEN Center for Advanced Intelligence Project (RIKEN AIP), Tokyo, Japan 2018Research InternIBM Research, Singapore 2015Undergraduate ExchangeUniversity of California, Los Angeles (UCLA) 2015Summer SchoolUniversity of Cambridge

Showing 5 of 9 entries · scroll within the list

Editorial Associate Editor, IEEE Transactions on Artificial Intelligence and Neural Networks

Conferences Area Chair, ICML, NeurIPS and ICLR · Senior Program Committee, AAAI

Research agenda

Four routes to safer, more dependable intelligence.

Each programme connects a hard scientific question to an observable failure mode—and to a path for intervention.

01

Foundation Model & Embodied AI Safety

How do we expose and control failure before intelligent agents reach the real world?

Safety evaluation and intervention for foundation models, multimodal systems, vision-language-action models, and agents with persistent memory.

Agent safetyVLA modelsRed teamingEvaluation
02

Adversarial Attacks, Data Poisoning & Robustness

What does reliable learning require when data or supervision can be actively corrupted?

Attacks, defenses, robust representation learning, and reliable training under adversarial examples, poisoned data, and noisy labels.

Adversarial learningData poisoningRobust trainingImperfect data
03

Reliable Robot Decision-Making & Risk Control

How should an embodied system act when its perception or environment is uncertain?

Uncertainty-aware perception, risk-sensitive policies, and fail-safe decision mechanisms for robots operating under distribution shift.

Robot learningUncertaintyOOD detectionRisk control
04

Trustworthy AI Benchmarks & Industry Solutions

How can scientific reliability become a repeatable engineering practice?

Reproducible benchmarks, stress tests, and deployment protocols that translate trustworthy-AI research into measurable system assurance.

BenchmarksAssuranceReproducibilityTechnology transfer

Open to strong prospective students

Prospective students & collaborators

Join RISE-AI and work on problems where reliability truly matters.

We welcome prospective master's students, direct-entry PhD students, research interns, and visiting collaborators who want to build rigorous and useful trustworthy AI.

“Excellent research grows from sustained effort, not last-minute sprints; from teamwork, not individual heroics.”— Jingfeng Zhang

Master's & direct-entry PhD

For students seeking deep training in trustworthy machine learning, embodied intelligence, or their intersection.

Research interns

For students ready to commit to a well-scoped research problem and a reproducible implementation.

Visiting researchers & collaborators

For complementary expertise in robotics, multimodal learning, AI safety, or application domains.

Who thrives here

  • You are curious about why a model fails—not only how to improve a score.
  • You can work independently, communicate clearly, and contribute to a team.
  • You value careful experiments, readable code, and honest negative results.
  • You are ready for long-horizon work with high standards and humane mentoring.

How to start

  1. Read the project themes and two or three recent papers that genuinely interest you.
  2. Prepare a concise CV, transcript, and links to code, papers, or substantial course projects.
  3. Email a short note explaining the problem you want to study, why it matters, and why RISE-AI is a fit.
Email subject[RISE-AI Application] Name — programme — intended year
Email your application

RISE-AI Lab

A rigorous, open, and collaborative lab.

RISE-AI combines high standards with academic freedom and warm, direct mentorship. We value rigorous reasoning, reproducible systems, intellectual generosity, and research that holds up beyond the benchmark.

Portrait of Jingfeng Zhang
Principal Investigator

Jingfeng Zhang

Trustworthy machine learning · safe embodied AI

Prospective member

Your work could begin here.

Tell us what difficult reliability problem you want to solve.

Start a conversation →

Selected publications

Research, organised by year.

This is a curated, identity-checked record. For the complete and most current list, visit Google Scholar ↗.

Talks & lectures

Ideas shared across institutions and communities.

Selected invited talks, lectures, and conference presentations on reliable and adversarial learning.

View the complete 43-item archive

Teaching & mentoring

Teaching that connects foundations to current systems.

Machine learning, generative modelling, trustworthy AI, and research mentoring across institutions in China, Japan, and New Zealand.

Current

Fudan University

Graduate Research Mentoring

Research supervision in trustworthy machine learning and safe embodied AI, grounded in rigorous process, academic freedom, and team collaboration.

Join RISE-AI

Students · collaborators · partners

Let’s make intelligent systems worthy of trust.

zjf@fudan.edu.cn

Room 508, Building D2, Bay Valley, Yangpu District, Shanghai, China

Scholar ↗ ORCID ↗ GitHub ↗