@inproceedings{huerta-enochian-ko-2024-instruction,
    title = "Instruction Fine-Tuning: Does Prompt Loss Matter?",
    author = "Huerta-Enochian, Mathew  and
      Ko, Seung Yong",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/jlcl-multiple-ingestion/2024.emnlp-main.1267/",
    doi = "10.18653/v1/2024.emnlp-main.1267",
    pages = "22771--22795",
    abstract = "We present a novel study analyzing the effects of various prompt loss token weights (PLW) for supervised instruction fine-tuning (SIFT). While prompt-masking (PLW = 0) is common for SIFT, some fine-tuning APIs support fractional PLWs and suggest that using a small non-zero PLW can help stabilize learning when fine-tuning on short-completion data. However, there has never been a study confirming this claim, and OpenAI, a major cloud-based SIFT provider, recently removed this parameter from their fine-tuning API. We found that performance of models fine-tuned on short-completion data had a statistically-significant negative quadratic relationship with PLW. Using small values (0.01 {\ensuremath{-}} 0.5) of PLW produced better results on multiple-choice and short-generation benchmarks (outperforming models fine-tuned on long-completion data) while large values ({\ensuremath{\approx}} 1.0) of PLW produced better results on long-generation benchmarks. We explained this effect and verified its importance through additional experiments. This research serves as a warning to API providers about the importance of providing a PLW parameter for SIFT."
}