让大模型记住自己之前的标注，性能提升5%到25%
Memory Is All You Need: Testing How Model Memory Affects LLM Performance in Annotation Tasks
- 让模型保留自身过往标注记录，实现上下文连续决策
- 相比零样本和少样本学习，准确率提升5%至25%
- 结合强化学习的记忆增强策略，在四次测试中三次表现更优
生成式大语言模型在零样本和少样本学习下已展现良好的文本标注能力。然而，这些方法不支持模型保留先前标注信息，导致每次响应彼此独立。本文研究模型记忆（即大模型对自身之前标注的认知）是否影响性能。基于OpenAI的GPT-4o与Meta的Llama 3.1，在两个政治科学数据集上实验表明，允许模型保留自身过往分类结果可带来显著性能提升：相比零样本和少样本学习，准确率提高5%至25%。此外，我们提出一种新的记忆增强方法，融合模型记忆与强化学习，在四次测试中有三次取得额外增益。该发现对希望提升大模型标注效率与效果的应用研究者具有重要价值。
原文摘要 · Abstract (English)
Generative Large Language Models (LLMs) have shown promising results in text annotation using zero-shot and few-shot learning. Yet these approaches do not allow the model to retain information from previous annotations, making each response independent from the preceding ones. This raises the question of whether model memory -- the LLM having knowledge about its own previous annotations in the same task -- affects performance. In this article, using OpenAI's GPT-4o and Meta's Llama 3.1 on two political science datasets, we demonstrate that allowing the model to retain information about its own previous classifications yields significant performance improvements: between 5 and 25\% when compared to zero-shot and few-shot learning. Moreover, memory reinforcement, a novel approach we propose that combines model memory and reinforcement learning, yields additional performance gains in three out of our four tests. These findings have important implications for applied researchers looking to improve performance and efficiency in LLM annotation tasks.
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。