Yefan Zhou's Personal Website

I am a CS PhD candidate (2023-) at Dartmouth College. I earned my Master's degree in EECS at UC Berkeley.
Previously, I interned at Salesforce AI Research.

Research: My research focuses on developing evaluation metrics and systems for ML models, spanning three dimensions: training quality, output reliability and hyperparameter. This research contributes to

Efficient training: LLM fine-tuning [4, 9, 10], hyperparameter optimization [7]
Auto-evaluation: LLM verification for reasoning [1], LLM-judge fine-tuning [3]
Inference optimization: pruning [5, 8], parallel decoding [2], ensembling [6]

News

⭐ Jan 2026: Three papers got accepted to ICLR 2026 (one 1st author, one Oral), thanks to my excellent collaborators and advisors! See you in Rio.

Jun 2025 - Sep 2025: I joined Salesforce AI Research as a Research Intern, check out our two ICLR papers on auto-evaluation ([1, 3]).

📣 Aug 2024: I passed the PhD qualification exam!

Recent Publications

First-author papers are highlighted and * indicates equal contribution

[1]	Variation in Verification: Understanding Verification Dynamics in Large Language Models Yefan Zhou, Austin Xu, Yilun Zhou, Janvijay Singh, Jiang Gui, Shafiq Joty ICLR 2026 Paper / Project / Twitter / Code [auto-evaluation, judge for reasoning, test-time scaling]
[2]	Diffusion Language Models Know the Answer Before Decoding {Pengxiang Li, Yefan Zhou}, Dilxat Muhtar, Lu Yin, Shilin Yan, Li Shen, Yi Liang, Soroush Vosoughi, Shiwei Liu ICLR 2026 Oral Paper / Code [efficient inference, diffusion language models, parallel decoding]
[3]	On the Shelf Life of Finetuned LLM-Judges: Future Proofing, Backward Compatibility, and Question Generalization Janvijay Singh, Austin Xu, Yilun Zhou, Yefan Zhou, Dilek Hakkani-Tür, Shafiq Joty ICLR 2026 Paper [judge fine-tuning, preference optimization]
[4]	Temperature Balancing, Layer-wise Weight Analysis, and Neural Network Training {Yefan Zhou, Tianyu Pang}, Keqin Liu, Charles H. Martin, Michael Mahoney, Yaoqing Yang NeurIPS 2023 Spotlight Paper / Code / Video [efficient training, NN optimizer, weight/layer analysis]
[5]	AlphaPruning: Using Heavy-Tailed Self Regularization Theory for Improved Layer-wise Pruning of Large Language Models {Haiquan Lu, Yefan Zhou}, Shiwei Liu, Zhangyang Wang, Michael W. Mahoney, Yaoqing Yang NeurIPS 2024 Paper / Code [efficient inference, LLM pruning, weight/layer analysis]
[6]	Sharpness-diversity tradeoff: improving flat ensembles with SharpBalance {Haiquan Lu, Xiaotian Liu, Yefan Zhou, Qunli Li}, Kurt Keutzer, Michael W. Mahoney, Yujun Yan, Huanrui Yang, Yaoqing Yang NeurIPS 2024 Paper / Code [Ensembling, Data selection, OOD]
[7]	MD tree: a model-diagnostic tree grown on loss landscape {Yefan Zhou, Jianlong Chen}, Qinxue Cao, Konstantin Schürholt, Yaoqing Yang ICML 2024 Paper / Code / Video [Scaling law, Hyperparameter tuning for training]
[8]	A Three-regime model of Network Pruning Yefan Zhou, Yaoqing Yang, Arin Chang, Michael Mahoney ICML 2023 Paper / Code / Video [NN pruning, efficient inference]
[9]	Model Balancing Helps Low-data Training and Fine-tuning Zihang Liu, Yuanzhe Hu, Tianyu Pang, Yefan Zhou, Pu Ren, Yaoqing Yang EMNLP main 2024 Oral Paper / Code [LLM fine-tuning, Layer quality analysis]
[10]	AlphaExpert: Assigning LoRA Experts Based on Layer Training Quality Peijun Qing, Chongyang Gao, Yefan Zhou, Xingjian Diao, Pu Ren, Yaoqing Yang, Soroush Vosoughi EMNLP main 2024 Paper [LLM efficient fine-tuning, Mixture-of-expert]

Work Experiences

Jun 2025 - Sep 2025: Research Intern @ Salesforce AI Research, working with Austin Xu, Yilun Zhou and Shafiq Joty.
Jan 2023 - Aug 2023: Research Intern @ UC Berkeley ICSI and Sky Computing Lab, working with Yaoqing Yang and Michael Mahoney.

Recent Talks

Nov. 2024: Talk at Snowflake, "Weight Matrix Diagnostics and Improved Large Language Model Compression and Fine-tuning"
Jan. 2024: Talk at AI-TIME, "Phase transition, loss landscape and model diagnostics"
Oct. 2023: Talk at UC Berkeley/ICSI TrojAI onsite, "Layer-wise Weight Analysis and Neural Network Training"
Mar. 2023: Talk at UC Berkeley/ICSI TrojAI onsite, "A Three-regime model of Network Pruning"

Last Updated: Oct 2025.

Website Template by Jon Barron