Fei Miao 苗菲

Endowed Associate Professor at School of Computing, University of Connecticut @Uconn, expert on Safe and Robust Embodied AI, PhD from UPenn GRASP @GRASPlab

[Fei Miao 苗菲 @fei_fay]

Exciting News: Paper Published in TMLR!
I am thrilled to announce that our research, "α-OCC: Uncertainty-Aware Camera-based 3D Semantic Occupancy Prediction," has been officially published in the Transactions on Machine Learning Research (TMLR)!
https://t.co/bSvQPHiJUS

[Fei Miao 苗菲 @fei_fay]

#Robotics #IROS #Keynote #IEEE_RAS Just found out that two months ago my IROS Keynote talk has been posted on Youtube: IROS 2025 Keynotes - Mechanisms and Controls: Fei Miao https://t.co/ZDzf3wM4UE via @YouTube

[Fei Miao 苗菲 @fei_fay]

#ICLR My student Yuan Zhuang will be presenting our paper at the morning poster session, April 23

[Guanya Shi @GuanyaShi]

I’m so tired of writing rebuttals to this kind of “lack of novelty” review: “This paper trivially combines A, B, and C, so the algorithmic novelty is limited.”

Technically, most (if not all) robotics papers are convex combinations of existing ideas.

I still deeply appreciate A+B+C papers—especially when they deliver:

- New capabilities: the “trivial combination” unlocks behaviors we simply couldn’t achieve before

- Sensible & organic design: A+B+C is clearly the right composition—not some arbitrary A′+B+C′

- Nontrivial interactions: careful analysis of the dynamics, coupling, or failure modes between A, B, C

- Rehabilitating old ideas: A was dismissed for years, but paired with modern B/C, it suddenly works—and teaches us why

- System-level & "interface" insight: the contribution is not any single piece, but how the pieces talk to each other

- Scaling laws or regimes: identifying when/why A+B+C works (and when it doesn’t)

- Engineering clarity: making something actually work robustly in the real world is not “trivial”

- New problem formulations: sometimes the real novelty is in the reformulation—only under this view does A+B+C make sense.

Maybe worth keeping these in mind when reviewing the next A+B+C paper : )

[Fei Miao 苗菲 @fei_fay]

@GuanyaShi Fully agreed!

[Fei Miao 苗菲 @fei_fay]

We’re excited about the potential of this framework and look forward to exploring its application in large-scale model pretraining.

[Fei Miao 苗菲 @fei_fay]

Our group's paper, "LD-MoLE: Learnable Dynamic Routing for Mixture of LoRA Experts," will appear at #ICLR 2026
📜 Paper: https://t.co/2q6Q2Wau3V
💻 GitHub: https://t.co/n4SetN0fLW
In this work we propose a fully differentiable dynamic routing framework for Mixture-of-LoRA-Experts

[Fei Miao 苗菲 @fei_fay]

At the core of LD-MoLE is a Sparsegen-based routing function with a closed-form solution, enabling the model to adaptively determine the number of experts to activate for each token at different layers. Our method demonstrates effective improvements in PEFT settings.

[Fei Miao 苗菲 @fei_fay]

Will be at Neurips, San Diego, Dec 2-7, look forward to meeting with friends. #NeurIPS2025

[Fei Miao 苗菲 @fei_fay]

#IROS25 Fantastic shows and performances at IROS, best conference experience ever:)

[Fei Miao 苗菲 @fei_fay]

#IROS25 #Keynote #SafeRL #RobustRL #MARL I'm honored to give a keynote speech at IROS'25 to share our research results and visions about safe and robust reinforcement learning/MARL as a promising training methodology for robots. The keynote talk videos will be online soon. https://t.co/IFNtKuwXgH.

[Fei Miao 苗菲 @fei_fay]

@SharonYixuanLi Briefly checked the paper and recommended to my students😃

[Fei Miao 苗菲 @fei_fay]

#ICRA #MARL #Robotics Our ICRA' 25 paper, "Safety Guaranteed Robust Multi-Agent Reinforcement Learning with Hierarchical Control for Connected and Automated Vehicles", opensource code: https://t.co/8S18ziGDT6; project website: https://t.co/RJv6X8urGh.
Check out to see how to train robust policies for multi-agent systems with safety guarantees!

[Fei Miao 苗菲 @fei_fay]

#OpenSource #IROS #MARL #Robotics Our group has opensource code for recent published papers of IROS 2025, we will present "YOLO-LLM: You Only LLM Once for Multi-agent Reinforcement Learning", check out the source code and play with it to see whether LLM is useful for MARL policy training: https://t.co/FOX3ZRElpN.

[Fei Miao 苗菲 @fei_fay]

#RSS2025 #Robotics #LLM #AI Exited to present for the full-day workshop “Large Foundation Model for Interactive Robot Learning” at RSS 2025! 🗓 Date: Wednesday, June 25📍 Location: Room SGM 123
🌐 Details: https://t.co/PMwe0wFGLF, my talk will be at: 12:00 - 12:20 Fei Miao (UConn) - Multi-agent Reinforcement Learning and LLM for Embodied AI