CollabEdit:迈向无损式协作式知识编辑
CollabEdit: Towards Non-destructive Collaborative Knowledge Editing
AI总结:
针对大语言模型协作知识编辑场景未被研究且存在知识重叠、冲突、遗忘三大挑战的问题,提出无损框架COLLABEDIT,通过新型模型合并机制模拟全局编辑并避免性能下降,实验验证其优于有损基线方法。
摘要:
大语言模型(LLM)的协作式学习已成为利用多方私有数据以保障效率与隐私的新范式。与此同时,针对大语言模型的知识编辑(KE)因能够显式调控大语言模型的行为也受到越来越多的关注,但协作式知识编辑场景(即多方的知识编辑结果以隐私保护且持续的方式进行聚合)尚未得到研究。为此,本文首次对协作式知识编辑展开深入探究,首先明确了该场景下特有的三大挑战:知识重叠、知识冲突和知识遗忘。随后我们提出了一种无损式协作知识编辑框架COLLABEDIT,该框架采用一种新颖的模型合并机制来模拟全局知识编辑行为,同时避免性能出现严重下降。在两个经典数据集上开展的大量实验表明,COLLABEDIT相较于其他有损基线方法性能更优,研究结果为解决协作式知识编辑的三大挑战及未来应用提供了思路。我们的代码可在https://github.com/LINs-lab/CollabEdit获取。
英文摘要:
Collaborative learning of large language models (LLMs) has emerged as a new paradigm for utilizing private data from different parties to guarantee efficiency and privacy. Meanwhile, Knowledge Editing (KE) for LLMs has also garnered increased attention due to its ability to manipulate the behaviors of LLMs explicitly, yet leaves the collaborative KE case (in which knowledge edits of multiple parties are aggregated in a privacy-preserving and continual manner) unexamined. To this end, this manuscript dives into the first investigation of collaborative KE, in which we start by carefully identifying the unique three challenges therein, including knowledge overlap, knowledge conflict, and knowledge forgetting. We then propose a non-destructive collaborative KE framework, COLLABEDIT, which employs a novel model merging mechanism to mimic the global KE behavior while preventing the severe performance drop. Extensive experiments on two canonical datasets demonstrate the superiority of COLLABEDIT compared to other destructive baselines, and results shed light on addressing three collaborative KE challenges and future applications. Our code is available at https://github.com/LINs-lab/CollabEdit.
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。