Publications and Reports

MixedNUTS: Training-Free Accuracy-Robustness Balance via Nonlinearly Mixed Classifiers,
Transactions on Machine Learning Research (TMLR), 2024.
Yatong Bai, Mo Zhou, Vishal M. Patel, and Somayeh Sojoudi.
Project code available at github.com/Bai-YT/MixedNUTS.
Results and models verified and listed on RobustBench leaderboard (second place for CIFAR-100 \(\ell_\infty\) as of Feb 2024).

Improving the Accuracy-Robustness Trade-Off of Classifiers via Adaptive Smoothing,
in SIAM Journal on Mathematics of Data Science (SIMODS) 6 (3), 2024.
Yatong Bai, Brendon G. Anderson, Aerin Kim, and Somayeh Sojoudi.
Project code available at github.com/Bai-YT/AdaptiveSmoothing.
Results and models verified and listed on RobustBench leaderboard (second place for CIFAR-100 \(\ell_\infty\) as of May 2023).

Efficient Global Optimization of Two-Layer ReLU Networks: Adversarial Training and Quadratic-time Algorithms,
in SIAM Journal on Mathematics of Data Science (SIMODS) 5 (2), 446-474, 2023.
Yatong Bai, Tanmay Gautam, and Somayeh Sojoudi.
2021 INFORMS Data Mining Best Student Paper Award Runner-Up Prize.

Ranking Manipulation for Conversational Search Engines,
in Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024.
Samuel Pfrommer*, Yatong Bai*, Tanmay Gautam, and Somayeh Sojoudi. *Equal contribution.
Oral presentation (top 2% of submissions, top 10% of accepted papers).
Dataset and collection pipeline available at huggingface.co/datasets/Bai-YT/RAGDOLL.
Experiment code available at github.com/spfrommer/cse-ranking-manipulation.
Poster at bai-yt.github.io/files/publications/RAGDOLL_Poster.pdf.

ConsistencyTTA: Accelerating Diffusion-Based Text-to-Audio Generation with Consistency Distillation,
in INTERSPEECH, 2024
Yatong Bai, Trung Dang, Dung Tran, Kazuhito Koishida, and Somayeh Sojoudi.
Live demo at huggingface.co/spaces/Bai-YT/ConsistencyTTA. Try it out!
Project website consistency-tta.github.io.
Poster at consistency-tta.github.io/poster.pdf. Project code available at github.com/Bai-YT/ConsistencyTTA.
Model checkpoints available at huggingface.co/Bai-YT/ConsistencyTTA.
Result presented on PapersWithCode.
Work done during internship at Microsoft (Applied Science Group).

Initial State Interventions for Deconfounded Imitation Learning,
in the IEEE Conference on Decision and Control (CDC), 2023.
Samuel Pfrommer, Yatong Bai, Hyunin Lee, Somayeh Sojoudi.

Mixing Classifiers to Alleviate the Accuracy-Robustness Trade-Off,
in Annual Learning for Dynamics & Control Conference (L4DC), 2024.
Yatong Bai, Brendon G. Anderson, and Somayeh Sojoudi.

Practical Convex Formulation of Robust One-Hidden-Layer Neural Network Training,
in American Control Conference (ACC), 2022.
Yatong Bai, Tanmay Gautam, Yu Gai, and Somayeh Sojoudi.

Let's Go Shopping (LGS) — Web-Scale Image-Text Dataset for Visual Concept Understanding,
ArXiv preprint arXiv:2401.04575, 2024.
Yatong Bai, Utsav Garg, Erhan Bas, Isidora Filipovic, Apaar Shanker, Haoming Zhang, Samyak Parajuli, Amelia N. Chu, Eugenia D Fomitcheva, Elliot Branson, Aerin Kim, Somayeh Sojoudi, and Kyunghyun Cho.
Work done during internship at Scale AI.

Imitation Human Behavior Model for Interactive Autonomous Driving Simulation.
Chenran Li, Akio Kodaira, Yatong Bai.
Course project for Fall 2021 CS 285 — Deep Reinforcement Learning (UC Berkeley).

Instrument Identification for Real-world Music Pieces.
Yatong Bai, Zhubo Zhou, Tianyi Zuo, Zhaoran Ma, Junzhe Ruan.
Course project for Summer 2020 CS 4641 — Machine Learning (Georgia Tech).
Project code available at github.com/Bai-YT/CS4641-Music-Instrument-Recognition.

Design and Implementation of a Data-Display MATLAB GUI for Cochlear Model Simulator.
Yatong Bai, 2018.