arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2412.17872cs.CLcs.AIcs.IR
收藏

面向信息丰富与概率提升的联合知识编辑

Joint Knowledge Editing for Information Enrichment and Probability Promotion

Renmin University of China(中国人民大学);Peking University(北京大学);Tencent(腾讯)

Wenhang Shi, Yiren Chen, Shuqing Bian, Xinyi Zhang, Zhe Zhao, Pengfei Hu, Wei Lu, Xiaoyong Du

AI总结:

本文提出基于对比的探测方法定位模型低层信息丰富与高层概率提升两个关键阶段,并开发联合知识编辑方法JEEP,在GPT-J和LLaMA等模型上验证了其有效性。

摘要:

存储在大型语言模型中的知识需要及时更新,以反映现实世界信息的动态特性。为了更新知识,大多数知识编辑方法集中在低层,因为最近对知识召回过程的探测揭示,答案信息在低层得到了丰富。然而,这些探测方法仅能且只能揭示原始答案的关键召回阶段,而编辑的目标是纠正模型对目标答案的预测。这种不一致性表明,探测方法及其相关的编辑方法均存在缺陷。为了缓解这种不一致性并确定关键编辑区域,我们提出了一种基于对比的探测方法,并定位了两个关键阶段,在这些阶段中,模型在原始答案和目标答案之间的行为出现分歧:低层的信息丰富和高层中的概率提升。基于这些见解,我们开发了面向信息丰富与概率提升的联合知识编辑(JEEP)方法,该方法联合编辑低层和高层以修改这两个关键召回阶段。考虑到双重修改带来的相互干扰和日益严重的遗忘问题,JEEP 的设计确保了对不同区域的更新共享相同的目标且互为补充。我们通过在多种模型(即 GPT-J (6B) 和 LLaMA (7B))上编辑多达数千个事实,并处理多样化的编辑目标(即添加事实性和反事实性知识),对 JEEP 进行了严格评估。在所有测试场景中,JEEP 均取得了最佳性能,验证了我们探测方法的发现和编辑方法设计的有效性。我们的代码和数据可在 https://github.com/Eric8932/JEEP 获取。

英文摘要:

Knowledge stored in large language models requires timely updates to reflect the dynamic nature of real-world information. To update the knowledge, most knowledge editing methods focus on the low layers, since recent probes into the knowledge recall process reveal that the answer information is enriched in low layers. However, these probes only and could only reveal critical recall stages for the original answers, while the goal of editing is to rectify model's prediction for the target answers. This inconsistency indicates that both the probe approaches and the associated editing methods are deficient. To mitigate the inconsistency and identify critical editing regions, we propose a contrast-based probe approach, and locate two crucial stages where the model behavior diverges between the original and target answers: Information Enrichment in low layers and Probability Promotion in high layers. Building upon the insights, we develop the Joint knowledge Editing for information Enrichment and probability Promotion (JEEP) method, which jointly edits both the low and high layers to modify the two critical recall stages. Considering the mutual interference and growing forgetting due to dual modifications, JEEP is designed to ensure that updates to distinct regions share the same objectives and are complementary. We rigorously evaluate JEEP by editing up to thousands of facts on various models, i.e., GPT-J (6B) and LLaMA (7B), and addressing diverse editing objectives, i.e., adding factual and counterfactual knowledge. In all tested scenarios, JEEP achieves best performances, validating the effectiveness of the revealings of our probe approach and the designs of our editing method. Our code and data are available at https://github.com/Eric8932/JEEP.

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。

↑