arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2503.09341cs.CLcs.ET
收藏

大语言模型检测有害计算术语的评估

An Evaluation of LLMs for Detecting Harmful Computing Terms

Clarksburg High School(克拉克斯堡高中);The Ohio State University(俄亥俄州立大学);Yale University(耶鲁大学);George Mason University(乔治梅森大学)

Joshua Jacas, Hana Winchester, Alicia Boyd, Brittany Johnson

AI总结:

本文通过评估多种编码器、解码器及编码器-解码器大语言模型在64个技术术语上的表现,发现解码器模型(如Gemini Flash 2.0和Claude AI)在有害语言语境分析上优于编码器模型(如BERT),为改进自动化检测工具提供了依据。

摘要:

在技术语境中检测有害和非包容性术语对于在计算领域营造包容性环境至关重要。本研究通过评估一个精心整理的技术术语数据库(每个术语都配有具体用例),探讨了模型架构对有害语言检测的影响。我们测试了一系列编码器、解码器和编码器-解码器语言模型,包括 BERT-base-uncased、RoBERTa large-mnli、Gemini Flash 1.5 和 2.0、GPT-4、Claude AI Sonnet 3.5、T5-large 和 BART-large-mnli。每个模型都使用标准化提示词来识别 64 个术语中的有害和非包容性语言。结果表明,解码器模型(尤其是 Gemini Flash 2.0 和 Claude AI)在细致入微的语境分析方面表现出色,而像 BERT 这样的编码器模型则展现出强大的模式识别能力,但在分类确定性方面存在困难。我们讨论了这些发现对改进自动化检测工具的意义,并强调了各模型在促进技术领域包容性沟通方面的特定优势和局限性。

英文摘要:

Detecting harmful and non-inclusive terminology in technical contexts is critical for fostering inclusive environments in computing. This study explores the impact of model architecture on harmful language detection by evaluating a curated database of technical terms, each paired with specific use cases. We tested a range of encoder, decoder, and encoder-decoder language models, including BERT-base-uncased, RoBERTa large-mnli, Gemini Flash 1.5 and 2.0, GPT-4, Claude AI Sonnet 3.5, T5-large, and BART-large-mnli. Each model was presented with a standardized prompt to identify harmful and non-inclusive language across 64 terms. Results reveal that decoder models, particularly Gemini Flash 2.0 and Claude AI, excel in nuanced contextual analysis, while encoder models like BERT exhibit strong pattern recognition but struggle with classification certainty. We discuss the implications of these findings for improving automated detection tools and highlight model-specific strengths and limitations in fostering inclusive communication in technical domains.

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。

↑