arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.00477cs.CL
收藏

意图自适应的LLM微调用于文本修订生成

Intention-Adaptive LLM Fine-Tuning for Text Revision Generation

Department of Computer Science, Learning Research & Development Center University of Pittsburgh(计算机科学系、学习研究与开发中心宾夕法尼亚大学)

Zhexiong Liu, Diane Litman

AI总结:

本文提出Intention-Tuning框架,通过意图自适应的逐层微调提升文本修订生成效果。

摘要:

大型语言模型(LLMs)在各种基于上下文的文本生成任务中展现了出色的性能,例如摘要和推理;然而,它们在基于意图的生成任务中的应用仍处于探索阶段。一个例子是修订生成,这要求生成的文本明确反映作者的实际意图。由于意图的复杂性和多样性,识别意图并生成理想的修订具有挑战性。尽管先前的工作利用LLMs通过少样本学习生成修订,但它们在处理交织的多意图场景时表现不佳。虽然使用基于意图的指令微调LLMs似乎有前途,但这种方法需要大量的标注数据,这在修订领域是昂贵且稀缺的。为了解决这些挑战,我们提出了Intention-Tuning,一种意图自适应的逐层LLM微调框架,该框架动态选择LLM的一部分层来学习意图,并将这些表示转移到修订生成。实验结果表明,Intention-Tuning在小规模修订语料库上有效且高效,优于几种PEFT基线。

英文摘要:

Large Language Models (LLMs) have achieved impressive capabilities in various context-based text generation tasks, such as summarization and reasoning; however, their applications in intention-based generation tasks remain underexplored. One such example is revision generation, which requires the generated text to explicitly reflect the writer's actual intentions. Identifying intentions and generating desirable revisions are challenging due to their complex and diverse nature. Although prior work has employed LLMs to generate revisions with few-shot learning, they struggle with handling entangled multi-intent scenarios. While fine-tuning LLMs using intention-based instructions appears promising, it demands large amounts of annotated data, which is expensive and scarce in the revision community. To address these challenges, we propose Intention-Tuning, an intention-adaptive layer-wise LLM fine-tuning framework that dynamically selects a subset of LLM layers to learn the intentions and subsequently transfers their representations to revision generation. Experimental results suggest that Intention-Tuning is effective and efficient on small revision corpora, outperforming several PEFT baselines.

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。

↑