About Me

Robin Luo

Computer Science PhD ∈ Northwestern University

I am a third-year Ph.D. candidate in the Computer Science Department at Northwestern University, advised by Professor Yan Chen and Han Liu in the MAGICS Lab.

My primary research interest lies in the intersection of Machine Learning, Health Care, and Financial Data.

Topics I am currently working on (Machine/Deep Learning):

  • Large Foundation Models:
    • Comprehensible Foundation Models, integrating modern Hopfield networks. [ICML ‘24]
    • Responsible Foundation Models, encompassing jailbreaks, adversarial attacks, and risks associated with scientific foundation models. [arXiv] [arXiv]
    • Accessible Foundation Models, incorporating PEFT, quantization, and outlier removal for efficient training and enhanced quantization robustness. [ICML ‘24] [ES-FoMo@ICML ‘24]
    • Actionable Foundation Models, featuring Chain of Thought and Chain of Action methodologies. [ICLR ‘25] [arXiv]
  • Applications of Large Foundation Models, including Genomic Foundation Models and Human Mobility Foundation Models. [arXiv] [HuMob@SIGSPATIAL ‘24]
  • Memory retrieval, memory-enhanced models, and memory editing techniques. [arXiv]

Topics I worked on in the past (before my Ph.D. studies):

Besides schoolwork and research, I have developed interests in many activities over time, including photography, hiking, and traveling.

News


Education

Northwestern University - MAGICS Lab

Computer Science PhD @ Northwestern University

Sept. 2022

Northwestern University

GPA 3.92/4.0, MSIT @ Northwestern University (Transfer to CS PhD program)

Sept. 2021

Georgia Institute of Technology

GPA 3.7/4.0, VQA-related research @ Machine Learning Lab

Augest. 2020 - May 2021
MS in Computer Science

Georgia Institute of Technology

Graduate with High Hornor, GPA 3.53/4.0, NLP-related research @ Machine Learning Lab, Dean List

Jan. 2018 - May 2020
BS in Computer Science

Michigan State University

Transfer to Georgia Institute of Technology with GPA 3.65/4.0

Sept. 2016 - Dec 2017
BS in Computer Engineering

My Research

Northwestern University

Integrated Modern Hopfield Networks into a knowledge-augmented language model to address limitations in self-attention architectures and foundation models, such as no-op outliers during training. Explored the interpretability of jailbreak strategies in safety-aligned LLMs by employing in-context learning and outlier analysis to enhance alignment and safety in superhuman AI systems. Applied LLMs to multivariate time-series challenges, including in-context time series forecasting, Limit Order Book dynamics prediction, and cross-sectional regression time-series analysis. Developed efficient hyperparameter optimization (HPO) methods for time-series prediction using the Tree-structured Parzen Estimator (TPE) and Genetic Algorithm.

Dec. 2021 - Present
Graduate Research Assistant

Northwestern University

Worked on distributed deep reinforcement learning by implementing the IQN model with the Wasserstein GAN algorithm. Designed a policy evaluation experiment using the Atari Game Database with fixed policy input and compared results with a baseline fixed policy. Developed policy optimization for the IQN model using the Wasserstein GAN algorithm.

Sept. 2021 - Nov. 2021
Graduate Research Assistant

Northwestern University

Worked on model-free reinforcement learning to achieve state-of-the-art performance in a physical simulator. Implemented PPO and SAC algorithms to compute and visualize reinforcement learning rewards. Contributed to a paper in progress, RobLAX-A Differentiable Robotics Framework for Physics-Augmented Reinforcement Learning.

May. 2021 - Sept. 2021
Graduate Research Assistant

Georgia Institute of Technology

Developed an NLP labeling pipeline for the brat annotation tool. Researched polysemy problems and applied the BERT model for target word and phrase labeling. Studied the Operational Transformation Algorithm to enable collaborative real-time editing in the cloud.

Sept. 2020 - May 2021
Graduate Research Assistant

Georgia Institute of Technology

Worked on Continuous Neuro-Symbolic Visual Question Answering (VQA), particularly researching soft logic functions. Designed and implemented a reasoning model combining BERT, bidirectional LSTM, and Stack-NMN, and evaluated its performance against other VQA models. Researched Video Question Answering and developed a model using R(2+1)D to detect object actions and movements. Contributed to a paper in progress, “Differentiable End-to-End Program Executor for Sample and Computationally Efficient VQA.”

Aug. 2020 - Oct 2020
Graduate Research Assistant

Georgia Institute of Technology

Researched the Visual Question Answering (VQA) problem using the NS-VQA algorithm to classify object relationships in images and answer GQA questions. Investigated reasoning mechanisms and improved performance using bidirectional LSTMs. Explored the functionality and latest advancements in VQA technology.

Aug. 2019 - May 2020
Undergraduate Research Assistant

Georgia Institute of Technology

Developed a computational pipeline for analyzing adsorption energy in chemical reactions using Python and the ASE API. Implemented a machine learning model to analyze chemical molecular structures.

Jan. 2019 - May 2020
DFT Model Analyze Adsorption Energies Team

Georgia Institute of Technology

Developed a data pipeline that collects and analyzes stadium network data to provide precise insights for stadium staff. Built a website tool for setting up a remote database for the LoPT Database Team.

Jan. 2018 - Dec 2018
VIP Georgia Tech – LoPt Database Team

Publication

A Spatial-Temporal Mixture-of-Experts Framework for Long-Term Cross-City Mobility Prediction

2nd ACM SIGSPATIAL International Workshop on the Human Mobility Prediction Challenge

Oct. 2024
Haoyu He, Haozheng Luo, Qi R. Wang

Decoupled Alignment for Robust Plug-and-Play Adaptation

Review by USENIX Security'2025

May. 2024
Haozheng Luo*, Jiahao Yu*, Wenxin Zhang*, Jialong Li, Jerry Yao-Chieh Hu, Xingyu Xin, Han Liu

Enhancing Jailbreak Attack Against Large Language Models through Silent Tokens

Review by USENIX Security'2025

May. 2024
Jiahao Yu*, Haozheng Luo*, Jerry Yao-Chieh Hu, Wenbo Guo, Han Liu, Xinyu Xing

Conv-CoA- Improving Open-domain Question Answering in Large Language Models via Conversational Chain-of-Action

Review by ICML'2025

May. 2024
Zhenyu Pan*, Haozheng Luo*, Manling Li, Han Liu

Fast Adaptation and Robust Quantization of Multi-Modal Foundation Models from Associative Memory - A Case Study in SpeechLM Authors

Accepted by Efficient Systems for Foundation Models II@ ICML2024

June. 2024
Shang Wu*, Yen-Ju Lu*, Haozheng Luo*, Jerry Yao-Chieh Hu, Jiayi Wang, Najim Dehak, Jesus Villalba, Han Liu

OutEffHop- A Principled Outlier-Efficient Attention Layer from Dense Associative Memory Models

Accepted by Efficient Systems for Foundation Models II@ ICML2024

June. 2024
Haozheng Luo*, Jerry Yao-Chieh Hu*, Pei-Hsuan Chang*, Hong-Yu Chen, Weijian Li, Wei-Po Wang, Han Liu

Outlier Efficient Modern Hopfield Model for Large Transformer-Based Models

Accepted by ICML'2024

Jan. 2024
Jerry Yao-Chieh Hu*, Pei-Hsuan Chang*, Haozheng Luo*, Hong-Yu Chen, Weijian Li, Wei-Po Wang, Han Liu

Chain-of-action- Faithful and Multimodal Question Answering through Large Language Models

Accepted by ICLR'2025

Jan. 2024
Zhenyu Pan, Haozheng Luo, Manling Li, Han Liu

SWGA - A Distributed Hyperparameter Search Method for Time Series Prediction Models

Review by ICML'2025

Aug. 2023
Weijian Li*, Haozheng Luo*, Chenwei Xu, Han Liu

SMUTF - Schema Matching Using Generative Tags and Hybrid Features

Review by Information Systems

Aug. 2023
Yu Zhang*, Di Mei*, Haozheng Luo*, Chenwei Xu, Richard Tzong-Han Tsai

IGN - Implicit Generative Networks

ICMLA'2022 - 2022 21st IEEE International Conference on Machine Learning and Applications

March. 2022
Haozheng Luo, Tianyi Wu, Feiyu Han, Zhijun Yan

RobLAX-A Differentiable Robotics Framework for Physics Augmented Reinforcement Learning

Sept. 2021
Guo Ye, Qinjie Lin, Tim Tsz-Kit Lau, Wanxin Jin, Cheng Zhou, Haozheng Luo, Zhuoran Yang, Zhaoran Wang, Han Liu

SciAnnotate- A Tool for Integrating Weak Labeling Sources for Sequence Labeling

Under Reviewed by ACL-SDProc 2024

Jan. 2021
Mengyang Liu *, Haozheng Luo*,Leonard Thong*, Yinhao Li, Chao Zhang, Le Song

Open-ended multi-modal relational reason for video question answering

RO-MAN'2023 - 2023 32nd IEEE International Conference on Robot and Human Interactive Communication

Dec. 2020
Haozheng Luo, Ruiyang Qin

HOW TO DESIGN SAMPLE AND COMPUTATIONALLY EFFICIENT VQA MODELS

Oct 2020
Karan Samel*, Zelin Zhao*, Kuan Wang*, Haozheng Luo, Binghong Chen, Song Le

Question ClassiNication with Deep Contextualized Transformer

FICC'2021 - Proceedings of the 2021 Future of Information and Communication Conference

Oct 2019
Haozheng Luo, Ningwei Liu, Charles Feng

My Career

Georgia Institute of Technology

I grade students’ projects and exams and provide feedback to facilitate improvement; Have a solid understanding of algorithms in KMM, Gradient Model and Neural-Networks; Provide students with the means to succeed by holding ofkice hours and responding to questions online

Augest. 2020 - May 2021
Graduate Teaching Assistant

Splunk

I automated the process of sending security alerts through Slack and Email by developing an app that would be deployed on all windows machines at Splunk; Developed a pipeline that detects sensitive kiles used by customer support and implements the required retention policies.

June. 2020 - Augest 2020
Backend Software Engineering Intern

Splunk

I designed and developed a customer support tool that automatically suggests answers to customer questions, by using Python/React.js/ Flask/Electron/TensorFlow and NLTK; Created a pipeline that updates the customer support knowledge base with customer support feedback from the previous tool, it potentially saved 35% of the original time.

May 2019 - Augest 2019
Backend Software Engineering Intern

Michigan Republican Party

I delivered a pipeline capable of predicting Twitter usernames from local residents' real names; Implemented a visualization tool that describes the analysis with several kinds of statistic chart

Augest 2018 - September 2018
Data Team Developer

R2.ai Inc.

I contributed work to the construction of a website tool which provides customers cloud-computing server; Designed a pipeline that tests machine learning functions which are inside web servers and compares results with common machine learning algorithms using Python and TensorFlow.

June 2018 - Augest 2018
Machine Learning Summer Intern

DBAPP Security

I developed a pipeline capable of scrawling data from the Common Vulnerability and Exposures website using Python and Selenium; Delivered the scrawled data into PostgreSQL

May 2017 - July 2017
Programmer Analyst Intern

Talks

  • Job Talk at Ryan Wang Lab
  • Funding Presentation with Ant Group
  • Oral Presentation at IEEE RO-MAN 2023
  • Group Reading Presentation on the Autoformer paper
  • Group Reading Presentation on the Gradient Boosting Algorithm
  • Oral Presentation at FICC 2021

Service

  • Teaching:
    • [Winter 2025] TA, Programming Massively Parallel Processors with CUDA (COMP_SCI 368), Northwestern University.
    • [Spring 2024] TA, Fundamentals of Computer Programming 1.5 (COMP_SCI 150), Northwestern University.
    • [Winter 2024] TA, Introduction to Artificial Intelligence (COMP_SCI 348), Northwestern University.
    • [Fall 2023] TA, Generative Methods (COMP_SCI 327), Northwestern University.
    • [Spring 2021] TA, Modeling and Simulation (CSE 6730), Georgia Institute of Technology.
    • [Fall 2020] TA, Computational Data Analysis (CSE 6740), Georgia Institute of Technology.
  • Reviewer: WWW 2020, NAACL 2024/2025, ACL 2020/2023/2024/2025, EMNLP 2023/2024, MLIS 2023, AIM 2024, NeurIPS 2024, ICLR 2025, ICML 2025, ACL-ARR (DOA).
  • Program Committee: EMNLP Industry Track 2023.

Collaboration and Mentoring

  • Chenghao Qiu, Tianjin University, BSCS '25 → CS Ph.D. study at TAMU (Fall '25)
  • Zhenyu Pan, MSECE '24 at the University of Rochester → CS Ph.D. study at NU (Fall '24)
    • Chain-of-Action: Faithful and Multimodal Question Answering through Large Language Models [ICLR '25]
    • Conv-CoA: Improving Open-Domain Question Answering in Large Language Models via Conversational Chain-of-Action [arXiv]
  • Mingyu Jin, MSCS '24 at NU → CS Ph.D. study at Rutgers University (Fall '24)
  • Hong-Yu Chen, NTU, Physics MS '24 → CS Ph.D. study at NU (Fall '24)
    • Outlier-Efficient Hopfield Layers for Large Transformer-Based Models [ICML '24]
  • Yihong Yu, University of California, Irvine, BSCS '25
  • Jingyu Elaine Wu, Hong Kong University of Science and Technology, BS '23 → Northwestern University, MSCS '25
  • Shaopeng Frank Gu, Northwestern University, CS + Statistics, BS '25

My Skills

My Projects

Snake Robot Simulator

Designed and developed a snake robot simulator using ROS API and Python.

...   ...

Worldwide Attacking & Protection Web Project

Developed a web application that visualizes global cybersecurity attacks using HTML5. Implemented a data pipeline to scrape cybersecurity attack data from the Kaspersky website and store it in a structured database.

...   ...

Object Relational REST API

Designed and implemented a RESTful API with PostgreSQL, enabling non-technical users to interact with the database easily. Developed a React-based frontend to showcase API functionality.

...   ...

Chatbot

Designed and implemented a chatbot application using Rasa for an interactive clothing recommendation system.

...   ...

Reward & Patent

Patents

  • Movable infusion stand (201120560457X)
    Robin Luo
    December 2013 in Hangzhou

Other

  • High Hornor of Graduateions
    May 2020 in Atlanta

Miscellaneous

30-30 Project

Hopefully, before I turn 30 years old, I can: