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 2405.19299cs.CL
收藏

面向去偏生成的专家引导有毒词元消除

Expert-Guided Extinction of Toxic Tokens for Debiased Generation

The Hong Kong Polytechnic University(香港理工大学);University of Technology Sydney(悉尼科技大学);The Education University of Hong Kong(香港教育大学)

Xueyao Sun, Kaize Shi, Haoran Tang, Guandong Xu, Qing Li

AI总结:

本文提出EXPOSED方法,通过构建去偏专家暴露并抑制有毒词元,在无需大量无偏语料或多轮指令的情况下,显著降低大型语言模型的社会偏见并平衡公平性与生成性能。

摘要:

大型语言模型(LLMs)在生成过程中可能会引发社会偏见,尤其是在使用有毒提示进行推理时尤为明显。控制生成中的敏感属性在数据分布、泛化能力和效率方面面临诸多挑战。具体而言,微调和检索需要大量无偏语料库,而直接提示则需要精心挑选的指令来在多轮思考中纠正输出,但这在内存和推理延迟方面带来了挑战。在本研究中,我们提出了面向去偏生成的专家引导有毒词元消除(EXPOSED),以在无需上述要求的情况下消除LLMs的有害输出。EXPOSED基于丰富的有毒语料库构建去偏专家,以暴露并引出潜在危险词元。随后,它处理LLMs的输出,通过抑制和削弱有毒词元来构建公平分布。EXPOSED在三个LLM家族的公平性基准上进行了评估。大量实验表明,与其他基线方法相比,所提出的EXPOSED在平衡公平性和生成性能的同时,显著降低了潜在的社会偏见。

英文摘要:

Large language models (LLMs) can elicit social bias during generations, especially when inference with toxic prompts. Controlling the sensitive attributes in generation encounters challenges in data distribution, generalizability, and efficiency. Specifically, fine-tuning and retrieval demand extensive unbiased corpus, while direct prompting requires meticulously curated instructions for correcting the output in multiple rounds of thoughts but poses challenges on memory and inference latency. In this work, we propose the Expert-Guided Extinction of Toxic Tokens for Debiased Generation (EXPOSED) to eliminate the undesired harmful outputs for LLMs without the aforementioned requirements. EXPOSED constructs a debiasing expert based on the abundant toxic corpus to expose and elicit the potentially dangerous tokens. It then processes the output to the LLMs and constructs a fair distribution by suppressing and attenuating the toxic tokens. EXPOSED is evaluated on fairness benchmarks over three LLM families. Extensive experiments demonstrate that compared with other baselines, the proposed EXPOSED significantly reduces the potential social bias while balancing fairness and generation performance.

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。

↑