I am a PhD student at Northeastern University, advised by Prof. Weiyan Shi. I was a scholar in the MATS program, working on reasoning at Prof. Dawn Song’s Lab. I obtained my Master’s degree in Computer Science at UMass Amherst where I was lucky to be advised by Prof. Andrew McCallum and Prof. Hong Yu. I finished my undergraduate study in Computer Science at HKUST.

I am generally interested in understanding the mechanisms of AI models to improve and control them. I am currently working on post-training, especially on understanding and mitigating emergent behaviors. I am broadly interested in continual learning, reasoning and interp. Feel free to email me if you would like to collaborate.

Publications

2026

• Can Aha Moments Be Fake? Identifying True and Decorative Thinking Steps in Chain-of-Thought
  Jiachen Zhao*, Yiyou Sun*, Weiyan Shi, Dawn Song
  Preprint.
  [pdf]

2025

• LLMs Encode Harmfulness and Refusal Separately
  Jiachen Zhao, Jing Huang, Zhengxuan Wu, David Bau, Weiyan Shi
  Neurips 2025.
  [pdf]

2024

• Large Language Models are In-context Teachers for Knowledge Reasoning
  Jiachen Zhao, Zonghai Yao, Zhichao Yang, Hong Yu
  EMNLP 24 Findings.
  [pdf]
  [TL;DR] We propose encoding specificity hypothesis inspired by humans’ memory retrieval to understand prompting LLM. We consider that effective prompts should match LLMs’ own training distribution.

• Learning and Forgetting Unsafe Examples in Large Language Models
  Jiachen Zhao, Zhun Deng, David Madras, James Zou, Mengye Ren
  ICML 24.
  [pdf]

• Multistage Collaborative Knowledge Distillation from a Large Language Model for Semi-Supervised Sequence Generation
  Jiachen Zhao, Wenlong Zhao*, Andrew Drozdov*, Benjamin Rozonoyer, Md Arafat Sultan, Jay-Yoon Lee, Mohit Iyyer, Andrew McCallum
  ACL 24.
  [pdf]
  

• Adaptive Fusion of Deep Learning with Statistical Anatomical Knowledge for Robust Patella Segmentation from CT Images
  Jiachen Zhao, Tianshu Jiang, Yi Lin, Justin Chan, Ping-Keung Lewis Chan, Chunyi Wen, Hao Chen
  Journal of Biomedical & Health Informatics.
  [pdf]
  

2023

• SELF-EXPLAIN: Teaching Large Language Models to Reason Complex Questions by Themselves
  Jiachen Zhao, Zonghai Yao, Zhichao Yang, Hong Yu
  Workshop on robustness of zero/few-shot learning in foundation models @ NIPS 23.
  [pdf]
  

• Student as an Inherent Denoiser of Noisy Teacher
  Jiachen Zhao
  3rd Workshop on Efficient Natural Language and Speech Processing @ NIPS 23.
  [pdf]
  [TL;DR] We find model converges to clean labels faster during knowledge distillation. We thus leverage the early checkpoint to denoise teacher labels.

• In-Context Exemplars as Clues to Retrieving from Large Associative Memory
  Jiachen Zhao
  Neural Conversational AI @ ICML 23. Associative Memory & Hopfield Networks @ NIPS 23.
  [pdf]
  

2022

• Trigger-free Event Detection via Derangement Reading Comprehension
  Jiachen Zhao, Haiqin Yang
  arXiv.
  [pdf]
  

Academic service

Reviewer for ICML, Neurips, CoNLL, AAAI, ICLR