Yihe Deng
Yihe Deng

PhD Student

About Me

I’m currently a 4th-year Ph.D. student at Department of Computer Science, University of California, Los Angeles (UCLA), where I am very fortunate to be advised by Prof. Wei Wang. Previously I received my B.Sc. from Department of Math, UCLA, and M.Sc. from Department of Computer Science, UCLA. During that time, I’ve been an student researcher at UCLA-NLP group with Prof. Kai-Wei Chang.

My research interests focus on post-training for Large Language Models (LLMs). Specifically, I’m interested in aspects including alignment fine-tuning, self-training and hallucination problems. I also work on robustness and multi-modal learning.

Interests
  • Large Language Models (LLMs)
  • LLM Post-training / Self-training
  • Multi-modal Learning
Education
  • PhD Computer Science

    University of California, Los Angeles

  • MS Computer Science

    University of California, Los Angeles

  • BS Mathematics of Computation

    University of California, Los Angeles

📚 My Research

I’m currently working on LLM post-training and focusing on self-training methods for both pure text LLMs and multi-modal LLMs. I’m generally interested in alignment and hallucination reduction for LLMs. I try to blog about papers I read.

Please reach out to discuss and collaborate at yihedeng at g dot ucla dot edu😃

🎉 Recent News
(09/2024) Our papers “Enhancing Large Vision Language Models with Self-Training on Image Comprehension” and “GraphVis: Boosting LLMs with Visual Knowledge Graph Integration” got accepted to NeurIPS 2024!
Featured Publications
Recent Publications

*Equal Contribution

(2024). MIRAI: Evaluating LLM Agents for Event Forecasting. arXiv preprint arXiv:2407.01231.
(2024). Enhancing Large Vision Language Models with Self-Training on Image Comprehension. Advances in neural information processing systems (NeurIPS).
(2024). Mitigating object hallucination in large vision-language models via classifier-free guidance. arXiv preprint arXiv:2402.08680.
(2024). Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models. Forty-first International Conference on Machine Learning (ICML).
(2024). Risk Bounds of Accelerated SGD for Overparameterized Linear Regression. The Twelfth International Conference on Learning Representations (ICLR).
Invited Talks