NeurIPS · 2025 · Conference paper · Top venue

Mitigating Forgetting in LLM Fine-Tuning via Low-Perplexity Token Learning

Chao-Chung Wu, Zhi Rui Tam, Chieh-Yen Lin, Yun-Nung (Vivian) Chen, Shao-Hua Sun, Hung-yi Lee

Court of Appeal · National Taiwan University of Science and Technology · Apple (Israel) · National Taiwan University

Published in
Advances in Neural Information Processing Systems 38
Date
2025-01-01
Citations
17
arXiv
2501.14315
Cite
@inproceedings{wu2025mitigating,
  title = {Mitigating Forgetting in LLM Fine-Tuning via Low-Perplexity Token Learning},
  author = {Chao-Chung Wu and Zhi Rui Tam and Chieh-Yen Lin and Yun-Nung (Vivian) Chen and Shao-Hua Sun and Hung-yi Lee},
  year = {2025},
  booktitle = {Advances in Neural Information Processing Systems 38},
  eprint = {2501.14315},
  archivePrefix = {arXiv},
  doi = {10.52202/085713-0064},
  url = {https://arxiv.org/abs/2501.14315},
}