Haotian Sun
ML Ph.D. Student @ Georgia Tech
Education
Aug. 2023 — present
Ph.D. in Machine Learning
Georgia Institute of Technology, Atlanta, GA
Advisor:
Dr. Bo Dai and Dr. Chao Zhang
Research focus: reinforcement learning, foundation models for decision making
Aug. 2022 — May 2023
M.S. in Computational Science and Engineering
Georgia Institute of Technology, Atlanta, GA
Advisor:
Dr. Chao Zhang
GPA: 4.00/4.00
Sept. 2015 — Aug. 2017
Engineer's Degree (Diplôme d'Ingénieur)
CentraleSupélec, Paris, France
Studied abroad through a dual degree program and obtained a graduate degree while enrolled at XJTU.
GPA: 3.9/4.0
Sept. 2013 — Jun. 2020
Combined Bachelor's/Master's Degree in Electrical Engineering and Automation
Xi'an Jiaotong University (XJTU), Xi'an, China
Advisor:
Dr. Hao Yi
Enrolled in XJTU Honors Youth Program, a highly selective nationwide program accepting under 120 students each year.
GPA: 90.7/100
Research Experiences
Fall 2023
Graduate Research Assistant, Georgia Tech
- Proposed an effective adapting approach for Black-Box LLMs, which offers a transparent, privacy-conscious, and cost-effective solution for customizing commercial black-box LLMs with only APIs;
- Designed an online adaptation framework iteratively sampling from previous inferences and optimizing the backend lightweight adapter (up to 0.3B);
- Achieved 5.90% improvement over the base model with 31.30 times less training cost and 1.84 times less inference cost than the official SFT service.
- Designed an online adaptation framework iteratively sampling from previous inferences and optimizing the backend lightweight adapter (up to 0.3B);
- Achieved 5.90% improvement over the base model with 31.30 times less training cost and 1.84 times less inference cost than the official SFT service.
Spring 2023
Graduate Research Assistant, Georgia Tech
- Proposed a closed-loop approach, AdaPlanner, which allows the LLM agent to refine its self-generated plan adaptively in response to environmental feedback;
- Developed a code-style LLM prompt structure that facilitates plan generation across a variety of tasks, environments, and agent capabilities;
- Proposed a skill discovery mechanism that leverages successful plans as few-shot exemplars, enabling the agent to plan and refine with fewer task demonstrations.
- Developed a code-style LLM prompt structure that facilitates plan generation across a variety of tasks, environments, and agent capabilities;
- Proposed a skill discovery mechanism that leverages successful plans as few-shot exemplars, enabling the agent to plan and refine with fewer task demonstrations.
Spring 2023
Graduate Research Assistant, Georgia Tech
- Introduced a new dataset called ToolQA, which is designed to faithfully evaluate LLMs' ability to use external tools for question answering;
- Minimized the overlap between our benchmark data and LLMs' pre-training data, enabling a more precise evaluation of LLMs' tool-use reasoning abilities;
- Conducted an in-depth diagnosis of existing tool-use LLMs to highlight their strengths, weaknesses, and potential improvements.
- Minimized the overlap between our benchmark data and LLMs' pre-training data, enabling a more precise evaluation of LLMs' tool-use reasoning abilities;
- Conducted an in-depth diagnosis of existing tool-use LLMs to highlight their strengths, weaknesses, and potential improvements.
Fall 2022
Graduate Research Assistant, Georgia Tech
- Proposed a node-absorbing diffusion process that operates directly in the discrete graph space;
- Designed a diffusion ordering network that learns an optimal node absorbing ordering from graph topology and a denoising network that uses the reverse node order to reconstruct the graph efficiently;
- Achieved better generation performance than previous state-of-the-art and guaranteed fast generation speed.
- Designed a diffusion ordering network that learns an optimal node absorbing ordering from graph topology and a denoising network that uses the reverse node order to reconstruct the graph efficiently;
- Achieved better generation performance than previous state-of-the-art and guaranteed fast generation speed.
Publications
C8
Haotian Sun,
Yuchen Zhuang,
Wei Wei,
Chao Zhang,
Bo Dai
arXiv .
C7
Haotian Sun,
Yuchen Zhuang,
Lingkai Kong,
Bo Dai,
Chao Zhang
Accepted at NeurIPS 2023 .
C6
Yuchen Zhuang,
Yue Yu,
Kuan Wang,
Haotian Sun,
Chao Zhang
Accepted at NeurIPS 2023 (Datasets and Benchmarks Track) .
C5
Lingkai Kong,
Jiaming Cui,
Haotian Sun,
Yuchen Zhuang,
B Aditya Prakash,
Chao Zhang
Accepted at ICML 2023 .
J3
Haotian Sun,
Hao Yi,
Guangyu Yang, et al.
IEEE Access 2019.
J2
Haotian Sun,
Hao Yi,
Fang Zhuo, et al.
IEEE Transactions on Power Delivery 2020.
Honors and Awards
2020
Outstanding Graduate Award at Xi'an Jiaotong University
For top 5% graduates graduating in 2020.
2018 - 2020
Excellent Student Award at the XJTU Power Electronics and Renewable Energy Center
For graduate student with top 5% academic records.
2019
ZHUHAI ENPOWER ELECTRIC Scholarship
For graduate student with top 10% academic records.
2018 - 2019
Excellent Academic Scholarship of Xi'an Jiaotong University
For graduate student with top 10% academic records.
2018
Excellent Academic Scholarship of Xi'an Jiaotong University
For graduate student with top 10% academic records.
2017
Tengfei Cup National College Student Business Plan Competition
Bronze Award
2013 - 2014
Pengkang Scholarship of Xi'an Jiaotong University
For graduate student with top 10% academic records.