Me
Hefan Zhang
PhD Candidate, Dartmouth College
hefan.zhang.gr [at] dartmouth.edu
Another me
I enjoy Chinese Calligraphy, playing Pipa, and have a strong appreciation for logic-based puzzle games that emphasize mathematical reasoning, such as Sudoku.

About

I am a Ph.D. candidate in Computer Science at Dartmouth College, where I am fortunate to be advised by Soroush Vosoughi. My research interests lie in natural language processing and machine learning, with a focus on LLM interpretability, reliability, and generation, as well as post-training LLM agents for long-horizon tasks. I'm always looking for related colaboration and internship opportunities. If you are interested to chat with me, feel free to drop me an email.

Interests

Education

Dartmouth College
Shanghai Jiao Tong University
  • Bachelor in Information Engineering and Mathematics Sep. 2018 – June 2022

News

  • Served as a reviewer for ICLR 2027.
  • Our paper Harbor Adaptor was accepted to NeurIPS 2026!
  • Our paper on uncertainty measurement was accepted to Findings of EMNLP 2026!
  • Started a research internship at Meituan.

Experiences

Jun. 2026 – Sep. 2026
Meituan
Large Language Model Research Intern
Post-training LLM agents for long-horizon tool-use tasks.
Sep. 2023 – Present
Dartmouth College
Ph.D. Candidate in Computer Science
Advisor: Soroush Vosoughi
Sep. 2018 – Jun. 2022
Shanghai Jiao Tong University (SJTU)
B.Eng. in Information Engineering
Minor in Applied Mathematics

Publications ( / )

SciPiVot benchmark construction and evaluation overview
SciΠvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents
Youran Sun, Xingyu Ren, Chugang Yi, Kejia Zhang, Di Wu, Yingkang Cao, Allen Zhou, Jianping Ye, Suying Liu, Jiaxuan Guo, Jianda Du, Weicheng Ye, Kecheng Liu, Hefan Zhang, Haixin Wang, Jie Xiao, Eric Yang, Haizhao Yang
arXiv 2026
[Paper] [Code]
Harbor Adapters infrastructure connecting benchmarks and agents
Harbor Adapters and Harbor-Index: Infrastructure and a Curated Meta-Dataset for Large-Scale Agentic Evaluation
Lin Shi, Haowei Lin, Zixuan Zhu, …, Hefan Zhang, et al.
NeurIPS 2026
[Paper] [Code]
Comparison of linguistic and internal confidence in large language models
When Linguistic and Internal Confidence Diverge in Large Language Models
Hefan Zhang, Bingquan Zhang, Ming Cheng, Saeed Hassanpour, Weicheng Ma, Soroush Vosoughi
EMNLP 2026 (findings)
[Paper] [Code]
Interpolation and extrapolation model merging overview
The Pitfalls of Text Degeneration when Aligning LLMs through Model Merge
Peijun Qing, Hefan Zhang, Lei Hsiung, Haiquan Lu, Xingjian Diao, Chiyu Ma, Saeed Hassanpour, Soroush Vosoughi
TMLR 2026
[Paper]
Enhancing LLM-Based Persuasion Simulations with Cultural and Speaker-Specific Information
Weicheng Ma*, Hefan Zhang*, Shiyu Ji, Farnoosh Hashemi, Qichao Wang, Ivory Yang, Joice Chen, Juanwen Pan, Michael Macy, Saeed Hassanpour, Soroush Vosoughi
EMNLP 2025 (findings)
[Paper] [Code]
A Generalizable Rhetorical Strategy Annotation Model Using LLM-based Debate Simulation and Labelling
Shiyu Ji*, Farnoosh Hashemi*, Joice Chen, Juanwen Pan, Weicheng Ma, Hefan Zhang, Sophia Pan, Ming Cheng, Shubham Mohole, Saeed Hassanpour, Soroush Vosoughi, Michael Macy
EMNLP 2025 (findings)
[Paper]
Recontextualizing revitalization: A mixed media approach to reviving the nüshu language
Ivory Yang, Xiaobo Guo, Yuxin Wang, Hefan Zhang, Yaning Jia, William Dinauer, Soroush Vosoughi
EMNLP 2025 (main)
[Paper]
Communication Makes Perfect: Persuasion Dataset Construction via Multi-LLM Communication
Weicheng Ma, Hefan Zhang, Ivory Yang, Shiyu Ji, Joice Chen, Farnoosh Hashemi, Shubham Mohole, Ethan Gearey, Michael Macy, Saeed Hassanpour, Soroush Vosoughi
NAACL 2025 (oral)
[Paper] [Code]
Efflex: Efficient and flexible pipeline for spatio-temporal trajectory graph modeling and representation learning
Ming Cheng, Ziyi Zhou, Bowen Zhang, Ziyu Wang, Jiaqi Gan, Ziang Ren, Weiqi Feng, Yi Lyu, Hefan Zhang, Xingjian Diao
CVPR 2024 Workshops
[Paper]
Improving Syntactic Probing Correctness and Robustness with Control Tasks
Weicheng Ma, Brian Wang, Hefan Zhang, Lili Wang, Rolando Coto-Solano, Saeed Hassanpour, Soroush Vosoughi
ACL 2023 (main)
[Paper]

GitHub

Benchmarking scientific judgment in long-horizon research agents.
Python
Star   - Fork   –
Unified benchmark adapters for large-scale evaluation of LLM agents.
Python
Star   - Fork   –
Measuring alignment between linguistic and internal confidence in LLMs.
Python
Star   - Fork   –
Persuasion Simulation with Culture Backgrounds by LLM agents.
Python
Star  1 Fork  –
Persuasion Dataset Generation by Multi-agent LLM framework.
Python
Star   - Fork   –

Selected Honors & Awards

  • First PrizeThe Chinese Mathematics Competitions 2019
  • Second PrizeChinese Mathematical Olympiad in Senior 2017
  • Individual Silver MedalWorld Mathematics Team Championship 2016
  • Second PrizeChinese Mathematical Olympiad in Senior 2016

Academic Services

Area Chair

  • ARRACL Rolling Review Oct. 2025

Peer Reviewer

  • ICLRInternational Conference on Learning Representations 2027
  • ACLAnnual Meeting of the Association for Computational Linguistics 2025, 2026
  • EMNLPConference on Empirical Methods in Natural Language Processing 2025, 2026
  • NLPCCNatural Language Processing and Chinese Computing 2026
Top