Kejian (Mark) Shi
mark.kejianshi[AT]gmail.com
The website is moving to https://kejianshi.com
I work on LLM posttraining, RL systems, and scalable oversight evaluation.
I had my MS research at Yale (23-25), advised by Arman Cohan at YaleNLP, and did my undergrad working with Sam Bowman at the NYU CS ARG group, where I still maintain active research involvement.
Publications
* denotes equal contributions. See Google Scholar for full list.
Sycophancy Towards Safety Researchers Causes Alignment Faking
In submission
Reference-Guided Self-Distillation: On-Policy Alignment for Non-Verifiable Domains
Kejian Shi*, Yixin Liu*, Peifeng Wang, Alexander Fabbri, Shafiq Joty, Arman Cohan
ICLR 2026 Main
SciRIFF: A Resource to Enhance Language Model Instruction-Following over Scientific Literature
PDF
Code
HF
Kejian Shi*, David Wadden*, Jacob Morrison, Aakanksha Naik, Shruti Singh, Nitzan Barzilay, Kyle Lo, Tom Hope, Luca Soldaini, Shannon Zejiang Shen, Doug Downey, Hannaneh Hajishirzi, Arman Cohan
EMNLP 2025 Main, NIPS 2024 Workshop FM4Science
ReIFE: Re-evaluating Instruction-Following Evaluation
PDF
Code
HF
Yixin Liu*, Kejian Shi*, Alexander R. Fabbri, Yilun Zhao, Peifeng Wang, Chien-Sheng Wu, Shafiq Joty, Arman Cohan
NAACL 2025 Main
Pretraining Language Models with Human Preferences
PDF
Tomasz Korbak, Kejian Shi, Angelica Chen, Rasika Bhalerao, Christopher L. Buckley, Jason Phang, Samuel R. Bowman, Ethan Perez
ICML'23: Proceedings of the 40th International Conference on Machine Learning
Oral Presentation
Sycophancy Towards Safety Researchers Causes Alignment Faking
In submission
Reference-Guided Self-Distillation: On-Policy Alignment for Non-Verifiable Domains
Kejian Shi*, Yixin Liu*, Peifeng Wang, Alexander Fabbri, Shafiq Joty, Arman Cohan
ICLR 2026 Main
SciRIFF: A Resource to Enhance Language Model Instruction-Following over Scientific Literature
PDF
Code
HF
Kejian Shi*, David Wadden*, Jacob Morrison, Aakanksha Naik, Shruti Singh, Nitzan Barzilay, Kyle Lo, Tom Hope, Luca Soldaini, Shannon Zejiang Shen, Doug Downey, Hannaneh Hajishirzi, Arman Cohan
EMNLP 2025 Main, NIPS 2024 Workshop FM4Science
ReIFE: Re-evaluating Instruction-Following Evaluation
PDF
Code
HF
Yixin Liu*, Kejian Shi*, Alexander R. Fabbri, Yilun Zhao, Peifeng Wang, Chien-Sheng Wu, Shafiq Joty, Arman Cohan
NAACL 2025 Main
Pretraining Language Models with Human Preferences
PDF
Tomasz Korbak, Kejian Shi, Angelica Chen, Rasika Bhalerao, Christopher L. Buckley, Jason Phang, Samuel R. Bowman, Ethan Perez
ICML'23: Proceedings of the 40th International Conference on Machine Learning
Oral Presentation
People Who Have Empowered Me
- My family
- Prof Sam Bowman, Prof Arman Cohan, Dr. Tomasz Korbak, Yixin Liu, Rachel Li, among others
Acknowledgement
This website uses the website design and template by Rose Wang and Martin Saveski