arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.00510cs.CLcs.AI
收藏

ROME:从文本、日志和表示中获得的记忆洞察

ROME: Memorization Insights from Text, Logits and Representation

Bo Li, Qinghua Zhao, Lijie Wen

AI总结:

ROME研究通过分析日志和表示差异,揭示记忆机制,发现长词更难记忆,高置信度与记忆能力正相关。

摘要:

先前工作通过比较模型输出与训练语料评估记忆,但分析大规模语料耗时。本文提出ROME方法,通过选择三种类型数据集重新定义记忆为正确回答能力,分析生成文本的日志和表示差异,发现长词记忆率低,高置信度与记忆正相关,相同概念在不同上下文中的表示更相似。代码和数据将在论文接受后公开。

英文摘要:

Previous works have evaluated memorization by comparing model outputs with training corpora, examining how factors such as data duplication, model size, and prompt length influence memorization. However, analyzing these extensive training corpora is highly time-consuming. To address this challenge, this paper proposes an innovative approach named ROME that bypasses direct processing of the training data. Specifically, we select datasets categorized into three distinct types -- context-independent, conventional, and factual -- and redefine memorization as the ability to produce correct answers under these conditions. Our analysis then focuses on disparities between memorized and non-memorized samples by examining the logits and representations of generated texts. Experimental findings reveal that longer words are less likely to be memorized, higher confidence correlates with greater memorization, and representations of the same concepts are more similar across different contexts. Our code and data will be publicly available when the paper is accepted.

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。

↑