arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.18680cs.CLcs.LG
收藏

非线性推理时干预:提升大语言模型的真实性

Non-Linear Inference Time Intervention: Improving LLM Truthfulness

University of Warsaw(华沙大学);Warsaw University of Technology(华沙理工大学)

Jakub Hoscilowicz, Adam Wiacek, Jan Chojnacki, Adam Cieslak, Leszek Michon, Vitalii Urbanevych, Artur Janicki

AI总结:

本文提出非线性推理时干预方法 NL-ITI,通过非线性多 token 探测与干预改进 LLM 真实性,在 TruthfulQA 上较基线 ITI 提升超 16% 相对 MC1。

摘要:

在这项工作中,我们探索了大语言模型(LLM)的内部表示空间,以识别包含最真实、最准确信息的注意力头。我们进一步开发了推理时干预(Inference Time Intervention,ITI)框架,该框架允许在不进行微调的情况下对 LLM 施加偏置。改进体现在引入了非线性多 token 探测和多 token 干预:非线性 ITI(NL-ITI),它显著提升了在评估基准上的性能。NL-ITI 在多个多项选择数据集上进行了测试,包括 TruthfulQA,在该数据集上我们报告了相对于基线 ITI 结果超过 16% 的相对 MC1(模型指向正确答案的准确率)提升。此外,相对于最近发布的同样专注于 ITI 改进的 Truth Forest(TrFf)方法,我们取得了 10% 的相对提升。

英文摘要:

In this work, we explore LLM's internal representation space to identify attention heads that contain the most truthful and accurate information. We further developed the Inference Time Intervention (ITI) framework, which lets bias LLM without the need for fine-tuning. The improvement manifests in introducing a non-linear multi-token probing and multi-token intervention: Non-Linear ITI (NL-ITI), which significantly enhances performance on evaluation benchmarks. NL-ITI is tested on diverse multiple-choice datasets, including TruthfulQA, on which we report over 16% relative MC1 (accuracy of model pointing to the correct answer) improvement with respect to the baseline ITI results. Moreover, we achieved a 10% relative improvement over the recently released Truth Forest (TrFf) method that also focused on ITI improvement.

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。

↑