PhD student, University of Southern California
1 paper at NeurIPS 2025
We empirically analyze, predict, and mitigate forgetting in upstream data during LLM fine-tuning.
