通过重构输入与优化输出，提升小数据场景下大模型表现
RIRO: Reshaping Inputs, Refining Outputs Unlocking the Potential of Large Language Models in Data-Scarce Contexts
- 双层架构：先用提示工程重写输入，再优化输出一致性
- Phi-2 在小数据上表现最佳，超越 Falcon 7B 与 Falcon 1B
- 适用于医疗、法律等高风险低数据领域
大语言模型在文本生成、摘要和问答等任务中表现卓越，但在小规模、特定领域的数据上微调时，常因泛化能力差而对陌生输入处理不佳。为此，我们提出 RIRO，一种两层架构：第一层利用先进提示工程重构输入，增强与训练数据的对齐；第二层聚焦输出精炼，减少不一致。在 Phi-2、Falcon 7B 与 Falcon 1B 上进行微调实验，结果显示 Phi-2 表现最优。我们还构建了包含余弦相似度、莱文斯坦距离、BLEU、ROUGE-1、ROUGE-2 与 ROUGE-L 等指标的评估基准。尽管性能有所提升，但计算开销与过拟合问题仍限制其在医疗、法律、软件测试等高风险低数据场景的应用潜力。
原文摘要 · Abstract (English)
Large language models (LLMs) have significantly advanced natural language processing, excelling in areas like text generation, summarization, and question-answering. Despite their capabilities, these models face challenges when fine-tuned on small, domain-specific datasets, often struggling to generalize and deliver accurate results with unfamiliar inputs. To tackle this issue, we introduce RIRO, a novel two-layer architecture designed to improve performance in data-scarce environments. The first layer leverages advanced prompt engineering to reformulate inputs, ensuring better alignment with training data, while the second layer focuses on refining outputs to minimize inconsistencies. Through fine-tuning models like Phi-2, Falcon 7B, and Falcon 1B, with Phi-2 outperforming the others. Additionally, we introduce a benchmark using evaluation metrics such as cosine similarity, Levenshtein distance, BLEU score, ROUGE-1, ROUGE-2, and ROUGE-L. While these advancements improve performance, challenges like computational demands and overfitting persist, limiting the potential of LLMs in data-scarce, high-stakes environments such as healthcare, legal documentation, and software testing.
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。