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 2311.09677cs.CL
收藏

R-Tuning:指示大语言模型说出“我不知道”

R-Tuning: Instructing Large Language Models to Say `I Don't Know'

University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);The Hong Kong University of Science and Technology(香港科技大学)

Hanning Zhang, Shizhe Diao, Yong Lin, Yi R. Fung, Qing Lian, Xingyao Wang, Yangyi Chen, Heng Ji, Tong Zhang

AI总结:

本文提出R-Tuning方法,通过基于参数知识与指令数据知识交集构建拒答感知数据来微调大语言模型,使其能拒绝回答超出自身知识范围的问题,从而缓解幻觉并提升不确定性校准能力。

摘要:

大语言模型(LLMs)凭借其出色的性能已经革新了众多领域,但仍面临诸多挑战。一个突出的问题是这些模型倾向于生成不存在的事实,这一问题被称为幻觉。我们的研究动机源于观察到以往的指令微调方法无论模型是否掌握相关知识,都会强制模型补全句子。当问题超出参数化知识范围时,模型会试图编造内容,并且无法表明自身缺乏相关知识。在本文中,我们提出了一种名为“拒答感知指令微调(R-Tuning)”的新方法。该方法首先识别预训练参数所包含的知识与指令微调数据所包含知识之间的差异,然后基于知识交集构建拒答感知数据,以微调LLMs,使其在面对超出自身参数化知识范围的问题时避免作答。实验结果表明,R-Tuning有效提升了模型回答已知问题的能力,同时增强了其对未知问题拒绝作答的能力。此外,在分布外数据集上的测试发现,拒答能力是一种可迁移至其他任务的元技能。进一步分析令人惊讶地发现,与基于不确定性的测试相比,学习不确定性能够带来更好的校准效果,并提升模型估计不确定性的能力。我们的代码可在 https://github.com/shizhediao/R-Tuning 获取。

英文摘要:

Large language models (LLMs) have revolutionized numerous domains with their impressive performance but still face their challenges. A predominant issue is the propensity for these models to generate non-existent facts, a concern termed hallucination. Our research is motivated by the observation that previous instruction tuning methods force the model to complete a sentence no matter whether the model knows the knowledge or not. When the question is out of the parametric knowledge, it will try to make up something and fail to indicate when it lacks knowledge. In this paper, we present a new approach called Refusal-Aware Instruction Tuning (R-Tuning). This approach is formalized by first identifying the disparity in knowledge encompassed by pre-trained parameters compared to that of instruction tuning data. Then, we construct the refusal-aware data based on the knowledge intersection, to tune LLMs to refrain from responding to questions beyond its parametric knowledge. Experimental results demonstrate R-Tuning effectively improves a model's ability to answer known questions and refrain from answering unknown questions. Furthermore, when tested on out-of-domain datasets, the refusal ability was found to be a meta-skill that could be generalized to other tasks. Further analysis surprisingly finds that learning the uncertainty results in better calibration and an improved ability to estimate the uncertainty than uncertainty-based testing. Our code is available at https://github.com/shizhediao/R-Tuning.

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。

↑