Associate Professor of ECE at CMU

Guannan Qu

I am broadly interested in machine learning, decision making, and control. My recent interest focuses on developing fundamental principles and scientific understandings of GenAI for interpretability, safety, and scalability. For a taste of this line of work, see here. I have also been interested in the interplay between learning and control, developing theories that make machine learning applicable in control of real-world large scale engineering systems. See here for example projects.

My work has been supported by multiple NSF grants, Air Force Office of Scientific Research, Amazon, Jane Street, C3 AI Institute, MFI, PITA (through a Bosch collaboration), Cylab, among others. My work has been recognized with an CAREER Award, AFOSR YIP, and multiple best (student) paper awards or finalists, including ECC 2026, ICRA 2025, AAAI 2025 Workshop on Multi-Agent AI in the Real World, IEEE Transactions on Smart Grid, among others.

My CV can be found here (updated in Aug 2026).

Note to Prospective PhD Students: I am looking to recruit one or more PhD students starting in Fall 2027. I am particularly interested in students who have either (1) genuine curiosity about scientifically understanding, interpreting, and improving modern AI models and algorithms (e.g., LLMs and physical AI); OR (2) a strong theoretical background and an interest in applying theory to modern AI.

Recent updates

Sept 2026PaperPaper on Effective Theory of LLMs, proposing a dynamical system perspective of LLM; see the project website for details and demos.
Aug 2026PaperPaper on GD-RHCR, a novel decentralized L-MAPF method motivated by our LIMDP theory.
Aug 2026PaperPaper on Activation-Keyed Momentum, a redesign of the momemtum used in common optimizers to cope with anisotropic features.
Aug 2026PaperPaper on Multi-Step Policy Gradient where we uncover and solve a surprising ``myopic`` local optimizer issue in policy optimization.
Aug 2026PaperPaper on Emergent Misalignment (EM) transfer where we systematically quantify EM transfer as a property of the dataset domain and task.
Aug 2026AwardReceived FY2026 Air Force Office of Scientific Research Young Investigator Program (AFOSR YIP) Award.
July 2026TalkGave an invited tutorial on dynamical systems perspectives of LLM in ECC 2026.
Jun 2026GrantReceived a new gift from Jane Street!
Jun 2026AwardPaper on transformers for networked control with long-range interactions has been selected for Best Student Paper Award for ECC 2026! Congratulations to our student authors, Vidur and Muhammed!
Jun 2026TalkInvited tutorial and Gen AI workshop talks at SIGMETRICS 2026.
Apr 2026PaperNew ICML paper on emergent misalignment.
Apr 2026GrantReceived a new grant from MFI.
Feb 2026NewsPromoted to Associate Professor of ECE (effective July 1, 2026).
Jan 2026PaperNew ICLR 2026 papers on internal planning in LLMs, generative control, and Riemannian diffusion.
Dec 2025TalkGave a tutorial at CDC on sampling-based control.
Dec 2025PaperNew NeurIPS 2025 papers on mean-field MARL (Spotlight) and stabilizing linear systems.
Aug 2025GrantReceived a new NSF grant on sampling-based control.
Aug 2025GrantReceived a new PITA grant.
Jul 2025PaperNew ICML 2025 paper on theoretical study of (hyper) self-attention.
May 2025AwardDial-MPC selected as a Best Paper Finalist at ICRA 2025.
Feb 2025AwardBest Paper Award at the AAAI 2025 Workshop on Multi-Agent AI in the Real World.
Feb 2025AwardGroup members received multiple fellowships: Zeji Yi (CMU Wei Shen and Xuehong Zhang Presidential Fellowship), Alex DeWeese (David H. Barakat and LaVerne Owen-Barakat College of Engineering Dean's Fellowship), and Chaoyi Pan (Hsu Chang Memorial Fellowship in Electrical & Computer Engineering).
Jan 2025ServiceCo-organized the NSF Workshop on Reinforcement Learning.
Past updates (2024 and earlier)
2024AwardStudent Alex DeWeese received the Leo Finzi Memorial Fellowship in Electrical & Computer Engineering.
2023AwardOur paper was selected among the top 5 papers (ranked 3rd) out of more than 1,000 articles published in IEEE Transactions on Smart Grid in the prior three years.
2023AwardReceived the NSF CAREER Award.
2023PaperPaper highlights:
  • Proposed CoVariance Optimal MPC (CoVO-MPC), exploiting dynamics structure to improve sampling-based MPC efficiency with strong theoretical and empirical results.
  • Applied our scalable RL framework from this paper to microgrid inverter control, demonstrating superior scalability; preprint here.
  • Proposed a distributed networked MPC framework with provable dynamic regret guarantees; preprint here.
  • Proposed an ISS-Lyapunov neural certificate framework to stabilize networked dynamical systems; accepted to L4DC 2023 (oral). See paper.
2023NewsThree new members joined the group (Fall 2023): Zeji, Chaoyi, and Muhammed.
2023GrantReceived CMU CyLab seed funding (Spring 2023).
2023PaperPaper "Global Convergence of Localized Policy Iteration in Networked Multi-Agent Reinforcement Learning" (link) accepted to SIGMETRICS 2023.
2023PaperPaper "Near-optimal distributed linear-quadratic regulator for networked systems" (link) accepted to SIAM Journal on Control and Optimization.
2022NewsTwo new members joined the group (Fall 2022): Alex and Ziyi.
2022PaperTwo papers accepted to NeurIPS 2022: On the sample complexity of stabilizing LTI systems on a single trajectory and Bounded-regret MPC via perturbation analysis: prediction error, constraints, and nonlinearity.
2022PaperOne paper in ICML 2022: Decentralized Online Convex Optimization in Networked Systems.
2022GrantReceived a new research grant from NSF EPCN (Spring 2022).
2022GrantReceived a new research award from C3 AI Institute (Spring 2022).
2021PaperPaper on scalable multi-agent RL for networked systems (link) accepted to Operations Research.
2021NewsStarted at CMU as an assistant professor (Fall 2021).

Flag Counter

Flag Counter