arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.11078cs.LGcs.AIcs.CL
收藏

通过标准微调进行模型编辑

Model Editing by Standard Fine-Tuning

Rutgers University(罗格斯大学)

Govind Gangadhar, Karl Stratos

AI总结:

本文提出通过对标准微调进行两处微小修改(优化条件似然及在未编辑事实上训练以鼓励局部性),使其在ZsRE和CounterFact数据集上的模型编辑性能匹配或优于高度专业化的编辑器。

摘要:

标准微调因其相对较差的性能,被认为不如模型编辑的专用方法有效。然而,它很简单,与被编辑模型的架构细节无关,并且能够直接利用标准训练技术的进步而无需额外工作(例如,用于计算效率的黑盒 PEFT),这使其成为模型编辑器的理想选择。在这项工作中,我们表明仅通过标准微调并进行两处微小修改,即可产生具有竞争力的模型编辑性能。首先,我们优化条件似然而非完全似然。其次,除了在随机改写的编辑提示上进行训练以鼓励泛化这一典型做法外,我们还在随机或相似的未编辑事实上进行训练以鼓励局部性。我们在 ZsRE 和 CounterFact 数据集上的实验表明,这些简单的修改使标准微调在编辑得分方面能够匹配或优于高度专业化的编辑器。

英文摘要:

Standard fine-tuning is considered not as effective as specialized methods for model editing due to its comparatively poor performance. However, it is simple, agnostic to the architectural details of the model being edited, and able to leverage advances in standard training techniques with no additional work (e.g., black-box PEFT for computational efficiency), making it an appealing choice for a model editor. In this work, we show that standard fine-tuning alone can yield competitive model editing performance with two minor modifications. First, we optimize the conditional likelihood rather than the full likelihood. Second, in addition to the typical practice of training on randomly paraphrased edit prompts to encourage generalization, we also train on random or similar unedited facts to encourage locality. Our experiments on the ZsRE and CounterFact datasets demonstrate that these simple modifications allow standard fine-tuning to match or outperform highly specialized editors in terms of edit score.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑