UMD Researchers Strong Showing with 24 Papers and an Outstanding Paper Award at ICML 2023

The paper titled "A Watermark for Large Language Models," stood out as one of only six outstanding papers from 1,865 accepted papers at the International Conference on Machine Learning 2023.
Descriptive image for UMD Researchers Strong Showing with 24 Papers and an Outstanding Paper Award  at ICML 2023

The University of Maryland (UMD) showcased its strong and notable presence at the prestigious 40th International Conference on Machine Learning (ICML 2023) held in Honolulu, Hawaii from July 23 to 29, with a strong focus on machine learning research.

UMD was well-represented by a diverse group of researchers, including faculty members, postdocs, and students, who collectively presented an impressive total of 24 papers and actively participated in 30 workshops.

Among the standout contributions from UMD researchers was the paper titled "A Watermark for Large Language Models," which received the Outstanding Paper Award. Spearheaded by the team of John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein, the study focuses on embedding unique and imperceptible identifiers, known as watermarks, into the large language model to help identify the source of specific outputs generated by the model. 

The breadth of UMD's research was truly impressive, covering a wide range of important topics in the field of machine learning. Some of these areas include measuring the robustness of neural networks, enhancing the accuracy of classifying images in unconventional settings, and exploring the potential of human assistance to aid AI agents in dealing with unfamiliar and challenging situations.

Overall, the significant presence and impactful research of UMD's machine learning community at ICML 2023 have further solidified the university's position as a leading institution in the field, making important contributions that advance the frontiers of machine learning and AI.

Below is the list of showcased papers from the University of Maryland (UMD):

Cramming: Training a Language Model on a single GPU in one day

 Jonas Geiping, Tom Goldstein (UMD) 

Text-To-Concept (and Back) via Cross-Model Alignment

 Mazda Moayeri, Keivan Rezaei, Maziar Sanjabi, Soheil Feizi (UMD) 

Continual Task Allocation in Meta-Policy Network via Sparse Prompting

Yijun Yang, Tianyi Zhou (UMD), Jing Jiang, Guodong Long, Yuhui Shi 

Learning Unforeseen Robustness from Out-of-distribution Data Using Equivariant Domain Translator

 Sicheng Zhu, Bang An, Furong Huang (UMD), Sanghyun Hong 

Identifying Interpretable Subspaces in Image Representations

  Neha Mukund Kalibhat, Shweta Bhardwaj, C. Bayan Bruss, Hamed Firooz, Maziar Sanjabi, Soheil   Feizi(UMD) 

Generalized Reductions: Making any Hierarchical Clustering Fair and Balanced with Low Cost

 Marina Knittel, Max Springer, John P Dickerson (UMD), MohammadTaghi Hajiaghayi (UMD)

Revisiting the Linear-Programming Framework for Offline RL with General Function Approximation 

 Asuman Ozdaglar, Sarath Pattathil, Jiawei Zhang, Kaiqing Zhang (UMD)

STEERING : Stein Information Directed Exploration for Model-Based Reinforcement Learning

 Souradip Chakraborty (UMD), Amrit Bedi (UMD), Alec Koppel, Mengdi Wang, Furong Huang (UMD), Dinesh Manocha (UMD)

A Watermark for Large Language Models

John Kirchenbauer(UMD), Jonas Geiping(UMD),Yuxin Wen, Jonathan Katz(UMD), Ian Miers, Tom Goldstein(UMD) 

GOAT: A Global Transformer on Large-scale Graphs

Kezhi Kong, Jiuhai Chen, John Kirchenbauer(UMD), Renkun Ni, C. Bayan Bruss, Tom Goldstein (UMD) 

Does Continual Learning Equally Forget All Parameters?

 Haiyan Zhao, Tianyi Zhou (UMD), Guodong Long, Jing Jiang, Chengqi Zhang

Structured Cooperative Learning with Graphical Model Priors

 Shuangtong Li, Tianyi Zhou (UMD), Xinmei Tian, Dacheng Tao 

Live in the Moment: Learning Dynamics Model Adapted to Evolving Policy

Xiyao wang, Wichayaporn Wongkamjan, Ruonan Jia, Furong Huang (UMD) 

Run-off Election: Improved Provable Defense against Data Poisoning Attacks

Keivan Rezaei, Kiarash Banihashem, Atoosa Malemir Chegini, Soheil Feizi (UMD)

Robust Counterfactual Explanations for Neural Networks With Probabilistic Guarantees

 Faisal Hamman, Erfaun Noorani, Saumitra Mishra, Daniele Magazzeni, Sanghamitra Dutta (UMD)

Auxiliary Modality Learning with Generalized Curriculum Distillation

 Yu Shen, Xijun Wang, Peng Gao, Ming Lin (UMD) 

Beyond Exponentially Fast Mixing in Average-Reward Reinforcement Learning via Multi-Level Monte Carlo Actor-Critic

 Wesley A. Suttle, Amrit Bedi (UMD), Bhrij Patel, Brian Sadler, Alec Koppel, Dinesh Manocha (UMD)

Partially Observable Multi-agent RL with (Quasi-)Efficiency: The Blessing of Information Sharing 

 Xiangyu Liu, Kaiqing Zhang (UMD)

Dynamic Constrained Submodular Optimization with Polylogarithmic Update Time

Kiarash Banihashem, Leyla Biabani, Samira Goudarzi, MohammadTaghi Hajiaghayi (UMD), Peyman Jabbarzade (UMD), Morteza Monemizadeh

Tighter Analysis for ProxSkip 

 Zhengmian Hu, Heng Huang (UMD)

Beyond Lipschitz Smoothness: A Tighter Analysis for Nonconvex Optimization 

 Zhengmian Hu, Xidong Wu,  Heng Huang (UMD)

A Law of Robustness beyond Isoperimetry 

 Yihan Wu, Heng Huang (UMD),  Hongyang Zhang

Fed-CBS: A Heterogeneity-Aware Client Sampling Mechanism for Federated Learning via Class-Imbalance Reduction

Jianyi Zhang, Ang Li (UMD), Minxue Tang, Jingwei Sun, Xiang Chen, Fan Zhang, Changyou Chen, Yiran Chen · Hai Li

Analyzing Convergence in Quantum Neural Networks: Deviations from Neural Tangent Kernels

Xuchen You, Shouvanik Chakrabarti, Boyang Chen, Xiaodi Wu (UMD)


—Article by Richa Mathur, CS Communications 


The Department welcomes comments, suggestions and corrections.  Send email to editor [-at-] cs [dot] umd [dot] edu.