用知识图谱指导模型持续学习，缓解领域迁移下的性能下降
Tackling Distribution Shift in LLM via KILO: Knowledge-Instructed Learning for Continual Adaptation
- 结合动态知识图谱与指令微调，实现持续适应新领域
- 在4个不同领域上均超越基线模型，保留旧知识并提升新任务表现
- 适合需要长期迭代更新的智能系统开发者
大型语言模型在面对领域迁移时常因灾难性遗忘导致性能下降。本文提出一种新的持续学习框架KILO（Knowledge-Instructed Learning for Continual Adaptation），将动态知识图谱与指令微调相结合。通过在训练中引入检索到的领域特定知识作为指导，KILO提升了模型对新领域的适应能力及对已有知识的保留。模型在WikiText-103上预训练，并在四个不同目标领域（BioASQ、SciQ、TweetEval、MIND）上评估顺序适应能力。实验表明，KILO在后向迁移、前向迁移、F1分数、保留率和训练效率等方面均持续优于强基线方法，包括持续微调、ERNIE 2.0和CPT。结果证明，结构化知识检索与指令提示相结合，能有效应对持续学习中的领域迁移挑战。
原文摘要 · Abstract (English)
Large Language Models (LLMs) often suffer from performance degradation when faced with domain shifts, primarily due to catastrophic forgetting. In this work, we propose KILO (Knowledge-Instructed Learning for Continual Adaptation), a novel continual learning framework that integrates dynamic knowledge graphs with instruction tuning. By leveraging retrieved domain-specific knowledge as guidance during training, KILO enhances both adaptability to new domains and retention of previously acquired knowledge. We pretrain our model on WikiText-103 and evaluate sequential adaptation across four diverse target domains: BioASQ, SciQ, TweetEval, and MIND. Our experiments demonstrate that KILO consistently outperforms strong baselines, including continual fine-tuning, ERNIE 2.0, and CPT, in terms of backward transfer, forward transfer, F1 score, retention rate, and training efficiency. These results highlight the effectiveness of combining structured knowledge retrieval and instruction prompting to overcome domain shift challenges in continual learning scenarios.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。