USC researchers will make a strong showing at ICML 2025, held July 13–19 in Vancouver, BC. Photo/USC
USC researchers will make a strong showing at ICML (the International Conference on Machine Learning) 2025, held July 13–19, with contributions across the poster, spotlight, and oral tracks. Their work reflects the breadth of innovation taking place across the university, spanning theoretical advances, health applications, and the frontiers of large language models. This year’s contributions include researchers from the USC Viterbi School of Engineering, the School of Advanced Computing’s Thomas Lord Department of Computer Science Ming Hsieh Department of Electrical and Computer Engineering, and the USC Marshall School of Business.
One prominent thread is the use of machine learning to better understand human behavior and improve health outcomes.
One prominent thread is the use of machine learning to better understand human behavior and improve health outcomes. From modeling behavioral signals in wearable data to interpreting neural imaging patterns, several papers focus on learning from complex biological and cognitive data. These efforts reflect a growing interest in merging data-driven methods with neuroscience, cognitive science and medicine.
On the technical front, USC-affiliated papers tackle core machine learning challenges, including optimization under uncertainty, causality, fairness, and the scalable training of large models. Researchers explored new strategies to enhance reasoning in language models, improve robustness in survival analysis, and enable vision-based reinforcement learning agents to generalize to novel environments. Together, these contributions underscore USC’s role in advancing both the science and real-world impact of machine learning.
Special thanks to Jing Yang, USC computer science student and founder of Paper Copilot, for assistance with this roundup.
USC-Affiliated Papers
(USC authors boldened)
Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions
Eray Erturk;Â Fahad Kamran;Â Salar Abbaspourazad;Â Sean Jewell;Â Harsh Sharma;Â Yujie Li;Â Sinead Williamson;Â Nicholas J Foti;Â Joseph Futoma
Session/area: applications->health medicine
Tightening Causal Bounds via Covariate-Aware Optimal Transport
Sirui Lin;Â Zijun Gao;Â Jose Blanchet;Â Peter Glynn;
Session/area: general machine learning->causality
Doubly Robust Conformalized Survival Analysis with Right-Censored Data (SPOTLIGHT)
Matteo Sesia;Â Vladimir Svetnik;
Session/area: probabilistic methods
Core Knowledge Deficits in Multi-Modal Language Models
Yijiang Li;Â Qingying Gao;Â Tianwei Zhao;Â Bingyang Wang;Â Haoran Sun;Â Haiyun Lyu;Â Robert D. Hawkins;Â Nuno Vasconcelos;Â Tal Golan;Â Dezhi Luo;Â Hokin Deng
Session/area: neuroscience, cognitive science
Fully Dynamic Euclidean Bi-Chromatic Matching in Sublinear Update Time (ORAL)
Gramoz Goranci;Â Peter Kiss;Â Neel Patel;Â Martin P. Seybold;Â Eva Szilagyi;Â Da Wei Zheng
Session/area: general machine learning
Poly2Vec: Polymorphic Fourier-Based Encoding of Geospatial Objects for GeoAI Applications
Maria Despoina Siampou;Â Jialiang Li;Â John Krumm;Â Cyrus Shahabi;Â Hua Lu;
Session/area: general machine learning
Jindong Tong;Â Hongcheng Liu;Â Johannes O. Royset;
Session/area: optimization > stochastic
Retraining with Predicted Hard Labels Provably Increases Model Accuracy
Rudrajit Das;Â Inderjit S Dhillon;Â Alessandro Epasto;Â Adel Javanmard;Â Jieming Mao;Â Vahab Mirrokni;Â Sujay Sanghavi;Â Peilin Zhong;
Session/area: theory->learning theory
Integer Programming for Generalized Causal Bootstrap Designs
Jennifer Rogers Brennan;Â Sebastien Lahaie;Â Adel Javanmard;Â Nick Doudchenko;Â Jean Pouget-Abadie;
Session/area: general machine learning->causality
Optimal Transport Barycenter via Nonconvex-Concave Minimax Optimization
Kaheon Kim;Â Rentian Yao;Â Changbo Zhu;Â Xiaohui Chen;
Session/area: optimization->nonconvex
Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM’s Reasoning Capability
Zicheng Lin;Â Tian Liang;Â Jiahao Xu;Â Qiuzhi Liu;Â Xing Wang;Â Ruilin Luo;Â Chufan Shi;Â Siheng Li;Â Yujiu Yang;Â Zhaopeng Tu;
Session/area: deep learning->large language models
Stochastic Control for Fine-tuning Diffusion Models: Optimality, Regularity, and Convergence
Yinbin Han;Â Meisam Razaviyayn;Â Renyuan Xu;
Session/area: deep learning->theory
Computing Voting Rules with Improvement Feedback
Evi Micha;Â Vasilis Varsamis
Session/area: theory->game theory
Asymmetric Decision-Making in Online Knowledge Distillation: Unifying Consensus and Divergence
Zhaowei Chen;Â Borui Zhao;Â Yuchen Ge;Â Yuhao Chen;Â Renjie Song;Â Jiajun Liang
Session/area: applications->computer vision
Smooth Interpolation for Improved Discrete Graph Generative Models
Yuxuan Song;Â Juntong Shi;Â Jingjing Gong;Â Minkai Xu;Â Stefano Ermon;Â Hao Zhou;Â Wei-Ying Ma
Session/area: deep learning->generative models and autoencoders
Improving the Variance of Differentially Private Randomized Experiments through Clustering
Adel Javanmard;Â Vahab Mirrokni;Â Jean Pouget-Abadie
Session/area: general machine learning->causality
Robust Conformal Outlier Detection under Contaminated Reference Data
Meshi Bashari;Â Matteo Sesia;Â Yaniv Romano
Session/area: probabilistic methods
Zero Shot Generalization of Vision-Based RL Without Data Augmentation
Sumeet Batra;Â Gaurav S. Sukhatme;
Session/area: reinforcement learning->deep rl
Test-Time Training Provably Improves Transformers as In-context Learners
Halil Alperen Gozeten;Â Muhammed Emrullah Ildiz;Â Xuechen Zhang;Â Mahdi Soltanolkotabi;Â Marco Mondelli;Â Samet Oymak;
Session/area: theory > deep learning
Dynamical Modeling of Behaviorally Relevant Spatiotemporal Patterns in Neural Imaging Data
Sayed Mohammad Hosseini;Â Maryam Shanechi
Session/area: applications->neuroscience cognitive science
Distilling the Knowledge in Data Pruning
Emanuel Ben Baruch;Â Adam Botach;Â Igor Kviatkovsky;Â Manoj Aggarwal;Â Gerard Medioni;
Session/area: deep learning->algorithms
DeepCrossAttention: Supercharging Transformer Residual Connections
Mike Heddes;Â Adel Javanmard;Â Kyriakos Axiotis;Â Gang Fu;Â Mohammadhossein Bateni;Â Vahab Mirrokni;
Session/area: deep learning->attention mechanisms
Diverging Preferences: When do Annotators Disagree and do Models Know?
Michael JQ Zhang;Â Zhilin Wang;Â Jena D. Hwang;Â Yi Dong;Â Olivier Delalleau;Â Yejin Choi;Â Eunsol Choi;Â Xiang Ren;Â Valentina Pyatkin
Session/area: deep learning->large language models
Muru Zhang;Â Mayank Mishra;Â Zhongzhu Zhou;Â William Brandon;Â Jue WANG;Â Yoon Kim;Â Jonathan Ragan-Kelley;Â Shuaiwen Leon Song;Â Ben Athiwaratkun;Â Tri Dao
Session/area: deep learning->large language models
End-to-End Learning Framework for Solving Non-Markovian Optimal Control
Xiaole Zhang;Â Peiyu Zhang;Â Xiongye Xiao;Â Shixuan Li;Â Vasileios Tzoumas;Â Vijay Gupta;Â Paul Bogdan;
Session/area: optimization
EARL-BO: Reinforcement Learning for Multi-Step Lookahead, High-Dimensional Bayesian Optimization
Mujin Cheon;Â Jay H Lee;Â Dong-Yeun Koh;Â Calvin Tsay
Session/area: optimization->zeroorder and blackbox optimization
Arya Fayyazi;Â Mehdi Kamal;Â Massoud Pedram
Session/area: social aspects->fairness
Contextual Linear Bandits with Delay as Payoff
Mengxiao Zhang;Â Yingfei Wang;Â Haipeng Luo
Session/area: theory->online learning and bandits
Synthetic Text Generation for Training Large Language Models via Gradient Matching
Dang Nguyen;Â Zeman Li;Â Mohammadhossein Bateni;Â Vahab Mirrokni;Â Meisam Razaviyayn;Â Baharan Mirzasoleiman
Session/area: deep learning->large language models
Position paper: The Artificial Intelligence and Machine Learning Community Should Adopt a More Transparent and Regulated Peer Review Process
Session/area: social ethical env impact
Note: Every effort was made to include all USC Viterbi-affiliated papers. If you believe your work was inadvertently left out, please let us know at cscomms@usc.edu so we can update the list.
Published on July 15th, 2025
Last updated on July 15th, 2025
