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 2406.14955cs.CL
收藏

ICLEval:评估大语言模型的上下文学习能力

ICLEval: Evaluating In-Context Learning Ability of Large Language Models

Gaoling School of Artificial Intelligence, Renmin University of China(中国人民大学高瓴人工智能学院);Huawei Poisson Lab(华为泊松实验室);School of Information, Renmin University of China(中国人民大学信息学院)

Wentong Chen, Yankai Lin, ZhenHao Zhou, HongYun Huang, Yantao Jia, Zhao Cao, Ji-Rong Wen

AI总结:

针对现有大模型评估框架忽略上下文学习能力测评的问题,本研究提出涵盖精确复制与规则学习两项子能力的ICLEval基准,证实该能力普遍存在且模型规模非唯一决定因素,还发现其在预训练早期即形成并稳定。

摘要:

上下文学习(In-Context Learning, ICL)是大语言模型(Large Language Models, LLMs)的一项核心能力,它使模型能够在相互关联的输入间进行理解与推理。评估LLMs的ICL能力既能提升其应用效果,也能加深我们对该能力在训练阶段形成机制的理解。然而,现有评估框架主要聚焦于语言能力与知识水平,往往忽略了对ICL能力的测评。本研究提出了用于评估LLMs ICL能力的ICLEval基准,该基准涵盖两项核心子能力:精确复制与规则学习。通过ICLEval基准我们证实,ICL能力普遍存在于不同的LLMs中,且模型规模并非ICL效果的唯一决定因素。令人意外的是,我们观察到ICL能力(尤其是复制能力)在预训练过程的早期就已形成,之后便趋于稳定。我们的源代码与基准已发布于https://github.com/yiye3/ICLEval。

英文摘要:

In-Context Learning (ICL) is a critical capability of Large Language Models (LLMs) as it empowers them to comprehend and reason across interconnected inputs. Evaluating the ICL ability of LLMs can enhance their utilization and deepen our understanding of how this ability is acquired at the training stage. However, existing evaluation frameworks primarily focus on language abilities and knowledge, often overlooking the assessment of ICL ability. In this work, we introduce the ICLEval benchmark to evaluate the ICL abilities of LLMs, which encompasses two key sub-abilities: exact copying and rule learning. Through the ICLEval benchmark, we demonstrate that ICL ability is universally present in different LLMs, and model size is not the sole determinant of ICL efficacy. Surprisingly, we observe that ICL abilities, particularly copying, develop early in the pretraining process and stabilize afterward. Our source codes and benchmark are released at https://github.com/yiye3/ICLEval.

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。

↑