Skip to main content
QUICK REVIEW

[论文解读] Using Natural Language Explanations to Rescale Human Judgments

Manya Wadhwa, Jifan Chen|arXiv (Cornell University)|May 24, 2023
Topic Modeling被引用 4
一句话总结

本文提出一种方法,利用自然语言解释和大语言模型(LLMs)将粗粒度的李克特量表人类判断重新缩放为细粒度的0–100分制评分,该方法受特定任务的评分标准指导。该方法在不降低标注者间一致性的前提下,提升了与专家判断的一致性,能够捕捉细微的主观差异和量表使用差异。

ABSTRACT

The rise of large language models (LLMs) has brought a critical need for high-quality human-labeled data, particularly for processes like human feedback and evaluation. A common practice is to label data via consensus annotation over human judgments. However, annotators' judgments for subjective tasks can differ in many ways: they may reflect different qualitative judgments about an example, and they may be mapped to a labeling scheme in different ways. We show that these nuances can be captured by natural language explanations, and propose a method to rescale ordinal annotations and explanations using LLMs. Specifically, we feed annotators' Likert ratings and corresponding explanations into an LLM and prompt it to produce a numeric score anchored in a scoring rubric. These scores should reflect the annotators' underlying assessments of the example. The rubric can be designed or modified after annotation, and include distinctions that may not have been known when the original error taxonomy was devised. We explore our technique in the context of rating system outputs for a document-grounded question answering task, where LLMs achieve near-human performance. Our method rescales the raw judgments without impacting agreement and brings the scores closer to human judgments grounded in the same scoring rubric.

研究动机与目标

  • 解决主观NLP任务中不一致且粗粒度的人类判断问题,特别是在评估大语言模型输出时。
  • 克服传统标签聚合和共识方法的局限性,这些方法无法捕捉细微的主观差异和量表使用差异。
  • 利用标注者提供的自然语言解释,提升数值评分的表达力,超越简单的李克特量表评分。
  • 开发一种后标注方法,利用大语言模型和可定制的评分标准,实现人类判断的细粒度重标度。
  • 提升高风险应用场景中用于训练和评估大语言模型的人类反馈的质量与保真度。

提出的方法

  • 在标注过程中收集每个判断的李克特量表评分(例如1–5)及相应的自然语言解释。
  • 通过提示词将每位标注者的李克特评分和解释输入大语言模型(GPT-4),提示词将重标度锚定于预定义的、基于维度的评分标准。
  • 利用大语言模型生成一个反映标注者潜在评估的0–100分制数值评分,该评分同时基于标签和解释内容。
  • 在后标注阶段基于对解释和判断的分析设计评分标准,从而实现原始标注方案中未编码的区分。
  • 通过将大语言模型的输出建立在结构化标准之上,确保重标度过程的一致性和可解释性,该标准为每个得分点定义了明确的评判标准。
  • 通过与参考人类判断对比评估重标度评分,以衡量其与专家级重标度的一致性与对齐程度。
Figure 1: Overview of our method. By feeding explanations that annotators write into an LLM, we can rescale their coarse-grained judgment to a 100-point scale.
Figure 1: Overview of our method. By feeding explanations that annotators write into an LLM, we can rescale their coarse-grained judgment to a 100-point scale.

实验结果

研究问题

  • RQ1基于自然语言解释的LLM重标度方法,能否生成与专家标注的0–100分制评分高度一致的评分?
  • RQ2所提出的方法是否在实现更细粒度判断区分的同时,保持了标注者间的一致性?
  • RQ3后标注评分标准在提升重标度评分保真度方面,相较于无标准指导的基线重标度方法,效果如何?
  • RQ4自然语言解释在多大程度上能够捕捉并传达在粗粒度李克特量表中丢失的细微主观感受?
  • RQ5该方法是否可推广至文档对齐问答之外的其他主观NLP评估任务?

主要发现

  • 所提出的基于解释的重标度(EBR)方法生成的评分,与专家标注的0–100分制评分相比,显著更接近基线重标度方法。
  • 重标度后,标注者间一致性(以克里波多夫的阿尔法系数衡量)保持稳定,表明该方法未扭曲原有的共识。
  • 使用评分标准显著提升了重标度的准确性,相较于无标准指导的重标度,凸显了结构化任务特定标准的价值。
  • 标注者的解释揭示了不同的推理模式,这些模式在重标度评分中得到了捕捉与体现,即使他们选择了相同的李克特标签。
  • 该方法成功区分了量表使用差异和主观性差异,实现了对标注者意图更忠实的表达。
  • 该方法在不同标注者之间具有鲁棒性,并保持了可解释性,适用于强化学习人类反馈(RLHF)及其他基于反馈的大语言模型对齐流程。
Figure 2: Example from Inquisitive-Broad where annotators were asked to label a question-answer pair on a Likert scale and give explanations for their judgments. Our proposed explanation-based rescaling (EBR) method maps these judgments using an aspect specific scoring rubric to scores that accurate
Figure 2: Example from Inquisitive-Broad where annotators were asked to label a question-answer pair on a Likert scale and give explanations for their judgments. Our proposed explanation-based rescaling (EBR) method maps these judgments using an aspect specific scoring rubric to scores that accurate

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。