Talks and Slides

Multi-step reasoning: Synthetic benchmarks, expressivity, and stability

• Talk slides at KDD’26 workshop

Second-order perspectives on fine-tuning, task attribution, RL, and beyond

• Talk slides: A Hessian view of supervised fine-tuning, task attribution, and reinforcement learning.

• Talk slides: In-context learning.

• Talk slides: A Hessian view of grokking.

• Guest lecture at Yale: “Generalization in Neural Networks: Recent Trend and Future Outlook,” March 2023.

• Talk at Northeastern CS Theory Lunch Seminar, March 2022.

Towards a foundation for multitask learning

• Slides: Boosting multitask learning. MSOM, June 2024.

• Slides: Recent work on multitask learning, March 2024.

• INFORMS: Transfer learning and random matrix theory, October 2022.

• One World ML Seminar: Understanding and improving generalization in multitask and transfer learning, May 2022.

• AAAI New Faculty Highlights: “Information Transfer in Multi-Task Learning, Data Augmentation, and Beyond,” February 2023.

• Colloquium talk at WPI CS Department and the ACM class at SJTU, November 2021.