Yongyuan Liang

Yongyuan (Cheryl) Liang's research focuses on developing foundation models and intelligent agents. She actively explores both theoretical frameworks and empirical findings, with specific research interests in:

Foundation models: Large multi-modal models/generative models for 2D/3D virtual and physical agentic tasks.
Alignment: Post-training alignment including human preference alignment and cross-modality alignment.
Trustworthy AI: Building robust and reliable agents for decision-making.

Email / Google Scholar / Github / Twitter

I join UMD CS as a PhD student, advised by Prof. Furong Huang. I received my B.S. degree in Mathematics from Sun Yat-sen University.

I'm always happy to collaborate with graduate/undergraduate students. Please drop me an email if you want to work with me.
I'm looking for part-time/full-time internship opportunities. Feel free to reach out if you're interested in my research.

Feb' 25	Magma to appear in CVPR 2025. Code and models have been released.
Jan' 25	Two papers to appear in ICLR 2025.
Jan' 25	Start to update Awesome-Generalist-Agents.
Sept' 24	Make-An-Agent to appear in NeurIPS 2024.
June' 24	ACE has been selected as a long oral presentation in ICML 2024.
May' 24	Two papers to appear in ICML 2024.
Feb' 24	Awarded a Dean's Fellowship.
Jan' 24	Three papers to appear in ICLR 2024, including two spotlights and one poster.

Selected Publications & Preprints

* denotes Equal Contributions and Project Lead; † indicates Equal Advising.

Foundation Model
	Avocado: A Unified Contrastive Framework for Multi-Objective Alignment of Language Models Coming sooooon, 2025 Project Page / Paper / Code / Models & Datasets / Twitter
	LEMON: A Unified and Scalable 3D Multimodal Model for Universal Spatial Understanding Coming soon, 2025 Project Page / Paper / Code / Models & Datasets / Twitter
	Magma: A Foundation Model for Multimodal AI Agents Magma Team CVPR, 2025 Project Page / Paper / Code / Models & Datasets / Twitter
	TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies Ruijie Zheng, Yongyuan Liang, Shuaiyi Huang, Jianfeng Gao, Hal Daumé III, Andrey Kolobov, Furong Huang, Jianwei Yang ICLR, 2025 ICLR Workshop GenBot, 2025(Oral Talks) Project Page / Paper / Code / Models / Twitter
	Robots Pre-Train Robots: Manipulation- Centric Robotic Representation from Large- Scale Robot Datasets Guangqi Jiang, Yifei Sun, Tao Huang, Huanyu Li, Yongyuan Liang†, Huazhe Xu† ICLR*, 2025 Project Page / Paper / Code / Models / Twitter
	Make-An-Agent: A Generalizable Policy Network Generator with Behavior-Prompted Diffusion Yongyuan Liang, Tingqiang Xu, Kaizhe Hu, Guangqi Jiang, Furong Huang, Huazhe Xu NeurIPS, 2024 NeurIPS Workshop AFM, 2024 (Oral Talks - Top 3%) Project Page / Paper / Code / Models & Dataset / Twitter
	PREMIER-TACO is a Few-Shot Policy Learner: Pretraining Multitask Representation via Temporal Action-Driven Contrastive Loss Ruijie Zheng, Yongyuan Liang, Xiyao Wang, Shuang Ma, Hal Daumé III, Huazhe Xu, John Langford, Praveen Palanisamy, Kalyan Basu, Furong Huang ICML, 2024 NeurIPS Workshop FMDM, 2023 Project Page / Paper / Code / Twitter
Decision-making
	ACE: Off-Policy Actor-Critic with Causality-Aware Entropy Regularization Tianying Ji, Yongyuan Liang, Yan Zeng, Yu Luo, Guowei Xu, Jiawei Guo, Ruijie Zheng, Furong Huang, Fuchun Sun, Huazhe Xu ICML, 2024 (Oral - Top 1.5%) Project Page / Paper / Code / Twitter
	DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization Guowei Xu, Ruijie Zheng, Yongyuan Liang, Xiyao Wang, Zhecheng Yuan, Tianying Ji, Yu Luo, Xiaoyu Liu, Jiaxin Yuan, Pu Hua, Shuzhen Li, Yanjie Ze, Hal Daumé III, Furong Huang, Huazhe Xu ICLR, 2024 (Spotlight - Top 5%)* CORL Workshop PRL, 2023 Project Page / Paper / Code / Twitter
Trustworthy AI
	Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations Yongyuan Liang, Yanchao Sun, Ruijie Zheng, Xiangyu Liu, Benjamin Eysenbach, Tuomas Sandholm, Furong Huang, Stephen Marcus McAleer ICLR, 2024 ICML Workshop AdvML-Frontiers, 2023 Paper / Code / Twitter
	Efficient Adversarial Training without Attacking: Worst-Case-Aware Robust Reinforcement Learning Yongyuan Liang, Yanchao Sun, Ruijie Zheng, Furong Huang NeurIPS, 2022 NeurIPS Workshop SafeRL, 2021 (Spotlight Talks) Paper / Code / Slides
	Is poisoning a real threat to LLM alignment? Maybe more so than you think Pankayaraj Pathmanathan, Souradip Chakraborty, Xiangyu Liu, Yongyuan Liang, Furong Huang AAAI, 2025 ICML Workshop on Models of Human Feedback for AI Alignment, 2024 Paper / Code
	Beyond Worst-case Attacks: Robust RL with Adaptive Defense via Non-dominated Policies Xiangyu Liu, Chenghao Deng, Yanchao Sun, Yongyuan Liang, Furong Huang ICLR, 2024 (Spotlight - Top 5%) NeurIPS Workshop MASEC, 2023 Project Page / Paper / Code / Twitter
	Who Is the Strongest Enemy? Towards Optimal and Efficient Evasion Attacks in Deep RL Yanchao Sun, Ruijie Zheng, Yongyuan Liang, Furong Huang ICLR, 2022 NeurIPS Workshop SafeRL, 2021 (Best Paper Award) Project Page / Paper / Code
	Certifiably Robust Policy Learning against Adversarial Communication in Multi-agent Systems Yanchao Sun, Ruijie Zheng, Parisa Hassanzadeh, Yongyuan Liang, Soheil Feizi, Sumitra Ganesh, Furong Huang ICLR, 2023 Paper / Code