Researcher

Minhao Cheng

Minhao Cheng is a machine-learning researcher whose work spans adversarial robustness, neural architecture search and trustworthy machine learning. He earned computer science degrees at the University of Electronic Science and Technology of China and UCLA, completing a PhD under Cho-Jui Hsieh in 2021. He coauthored the query-efficient Sign-OPT attack in 2020 and shared the 2021 ICLR Outstanding Paper Award for Rethinking Architecture Selection in Differentiable NAS. After serving at HKUST, he joined Penn State in Spring 2024, helped start TurningPoint AI, and coauthored recent work on retrieval and long-context language models; as of October 2026, he is an assistant professor in Penn State's College of Information Sciences and Technology.

Timeline

  1. 2015

    A computer science degree

    He earns a computer science degree at the University of Electronic Science and Technology of China in 2015.

  2. 2020

    A query-efficient attack

    He coauthors Sign-OPT, an ICLR 2020 method for hard-label black-box adversarial attacks that uses fewer model queries to generate adversarial examples.

  3. 2021

    A UCLA PhD and ICLR award

    He completes a computer science PhD at UCLA under Cho-Jui Hsieh in 2021. He shares the 2021 ICLR Outstanding Paper Award with Ruochen Wang, Xiangning Chen, Xiaocheng Tang and Cho-Jui Hsieh for Rethinking Architecture Selection in Differentiable NAS.

  4. 2022–2023

    An assistant professorship at HKUST

    He serves as an assistant professor in HKUST's Department of Computer Science and Engineering from 2022 to 2023. At HKUST's 2023 Research and Technology Forum, he presents work on post-breach protection for machine-learning security.

  5. 2024

    Joining Penn State IST

    He joins Penn State's College of Information Sciences and Technology as an assistant professor in Spring 2024.

  6. July 2024

    Starting TurningPoint AI

    He helps start TurningPoint AI in July 2024 as an AIGC research collaboration group across institutes.

  7. 2025–2026

    LaRA and Route Before Retrieve

    In 2025, he coauthors LaRA, a benchmark with 2,326 test cases across four question-answering task categories and three long-text types for comparing retrieval-augmented generation with long-context language models. In May 2026, he coauthors Route Before Retrieve, which uses document metadata to choose a retrieval or long-context strategy before answering.

  8. As of October 2026

    Machine learning at Penn State

    As of October 2026, he is an assistant professor in Penn State's College of Information Sciences and Technology. His research focuses on trustworthy machine learning and AutoML.

Career

  1. Since 2024Pennsylvania State UniversityAssistant Professor
  2. From 2024TurningPoint AIFounder
  3. 2022–2023Hong Kong University of Science and TechnologyAssistant Professor

Education

  1. 2021University of California, Los AngelesPhD
  2. 2015University of Electronic Science and Technology of ChinaComputer science degree

Sources and credits

Photo: We have no published photo of Minhao Cheng yet. Send us a free-licensed photo.

  1. Minhao Cheng - Minhao Cheng cmhcbb.github.io. Checked 21 October 2026.
  2. Sign-OPT: A Query-Efficient Hard-label Adversarial Attack for ICLR 2020 - IBM Research research.ibm.com. Checked 21 October 2026.
  3. Professor Hsieh and His Students Won the ICLR 2021 Outstanding Paper Award cs.ucla.edu. Checked 21 October 2026.
  4. Research and Technology Forum 2023 | HKUST CSE cse.hkust.edu.hk. Checked 21 October 2026.
  5. LaRA: Benchmarking Retrieval-Augmented Generation and Long-Context LLMs -- No Silver Bullet for LC or RAG Routing arxiv.org. Checked 21 October 2026.
  6. Route Before Retrieve: Activating Latent Routing Abilities of LLMs for RAG vs. Long-Context Selection arxiv.org. Checked 21 October 2026.