Scholar
Simin Fan
Google Scholar ID: YFJJxpQAAAAJ
EPFL
Citations & Impact
Citations
761
H-index
8
i10-index
8
Publications
17
Co-authors
10
list available
Publications
Resume (English only)
Academic Achievements
- 1. GRAPE: Optimize Data Mixture for Group Robust Multi-target Adaptive Pretraining;
- 2. NeuralGrok: Accelerate Grokking by Neural Gradient Transformation;
- 3. Dynamic Gradient Alignment for Online Data Mixing;
- 4. Task-Adaptive Pretrained Language Models via Clustered-Importance Sampling;
- 5. HyperINF: Unleashing the HyperPower of the Schulz's Method for Data Influence Estimation;
- 6. MEDITRON-70B: Scaling Medical Pretraining for Large Language Models;
- 7. DOGE: Domain Reweighting with Generalization Estimation;
- 8. Irreducible Curriculum for Language Model Pretraining;
- 9. ReadingQuizMaker: A Human-NLP Collaborative System that Supports Instructors to Design High-Quality Reading Quiz Questions;
- 10. Towards Process-Oriented, Modular, and Versatile Question Generation that Meets Educational Needs;
- 11. Genetic Risk Converges on Regulatory Networks Mediating Early Type-2 Diabetes;
- 12. Historical OCR Text Quality Analysis and Post-correction.
Research Experience
- 1. Conducting Ph.D. research at EPFL, focusing on the training of large foundation models;
- 2. Involved in multiple research projects, including data selection, efficient training algorithms, and model generalization.
Education
- 1. Ph.D. candidate in Machine Learning at École Polytechnique Fédérale de Lausanne (EPFL), advised by Prof. Martin Jaggi;
- 2. B.Sc. (honor) in Computer Science at University of Michigan, previously worked with Prof. Rada Mihalcea, Prof. Lu Wang, and Prof. Jie Liu;
- 3. B.Sc. (government honor) in Electrical and Computer Engineering at Shanghai Jiao Tong University.
Background
- Research interests include effective and efficient training of large foundation models, especially in data selection and curriculum design, efficient pretraining and post-training algorithms, understanding LLM training dynamics and generalization behaviors, and foundation models for scientific research (AI4Science).
Miscellany
- Hobbies include skiing, photography, piano, singing, ballet & yoga, tennis.
