首页 > AI前沿 > Enhancing Understanding of Language Models with Varied Explanations

Enhancing Understanding of Language Models with Varied Explanations

arXiv自然语言 2026-11-20 13:00 24 阅读 查看原文
用户需要我将一段纯文本新闻正文转换为语义化的HTML片段。我需要仔细分析要求: 1. 只输出HTML片段本身,不要Markdown围栏,也不要解释。 2. 每个自然段落用 `

` 包裹,长文用 `

`/`

` 划分小节。 3. 列表内容用 `
  • `,引用用 `
    `,代码用 `
    `,重要句子可用 ``。
    4. 保留原文所有信息,不得增删改内容。
    5. 不要输出 ``/``/`` 等文档级标签。
    
    现在来分析这段正文:
    
    这是一段关于语言模型可解释性的学术文本。让我看看它有怎样的逻辑结构:
    
    第一段(开头部分):介绍背景——语言模型用于支持人类判断,但其预测难以解释和验证。可解释性技术提供洞察。解释是模型内部工作的简化近似。目标是获得具有忠实性、合理性和效用的解释。评估方案很重要。
    
    第二段(中间部分):本论文研究高亮解释、自然语言解释和思维链推理。这些形式提供了互补的视角,在不同任务中表现出不同的优势和局限。生成满足多个质量维度的解释仍然具有挑战性。
    
    第三段(结尾部分):本论文旨在改善人们对模型决策过程的理解。首先,提出两个评估框架……其次,提出两种技术……最后,一项关于人机交互的研究……
    
    所以本质上可以分为三个部分:
    1. 背景与目标
    2. 研究范围与挑战
    3. 研究贡献(这实际上可以形成一个列表结构,因为描述了三个贡献,用“First”、“Second”、“Finally”引导)
    
    文本提到“First, it proposes... Second, it proposes... Finally, a study...”——这些自然地适合用列表呈现。
    
    我来安排结构:
    
    ```html
    

    背景与评估目标

    Language models are increasingly used...

    研究范围与挑战

    This thesis studies highlight explanations...

    研究贡献

    • First, it proposes two evaluation frameworks...
    • Second, it proposes two techniques...
    • Finally, a study of human-AI interaction...
    ``` 等等,我需要确保不增删改内容。原文中的“First, it proposes...”是句子的一部分,如果我把它拆成列表项,需要保持文字的完整性。让我再考虑一下。 实际上,原文是:“This thesis aims to improve people's understanding of models' decision-making processes in complex reasoning tasks. First, it proposes two evaluation frameworks for assessing highlight explanations across multiple properties and explanation types, together with a meta-evaluation framework for examining the reliability of existing faithfulness metrics for chain-of-thought explanations. Second, it proposes two techniques that generate natural language explanations designed to be faithful to the model's prediction process while remaining concise, coherent, and helpful for understanding model decisions. Finally, a study of human-AI interaction explores the roles of evidence and explanations in helping users validate model verdicts in a fact-checking setting.” 我需要保留所有信息。让我将“This thesis aims to...”这部分作为一个段落,然后将三个要点作为列表。但要注意,列表项需要是完整的句子,保持原文的措辞。 让我仔细重新构建结构,确保不丢失任何信息: 第一段(整体作为第一小节的背景介绍): “Language models are increasingly used to support human judgement in complex reasoning tasks, yet their predictions often remain difficult to interpret and validate. Explainability techniques provide insight into model decision-making, helping expose errors, failure modes, and potential bias so users can decide whether to rely on model outputs. Explanations are simplified approximations of a language model's internal workings. The goal is to obtain explanations that reflect the underlying computation as closely as possible (faithfulness), are understandable to users (plausibility), and help them achieve their intended goals (utility). Evaluation protocols are essential for assessing these dimensions.” 第二段: “This thesis studies highlight explanations, natural language explanations, and chain-of-thought reasoning. These forms offer complementary perspectives on model behaviour and exhibit distinct strengths and limitations across tasks. Producing explanations that satisfy multiple quality dimensions remains challenging, as does evaluating them reliably across forms and settings.” 第三部分(包含目标与三个贡献): “This thesis aims to improve people's understanding of models' decision-making processes in complex reasoning tasks.” 然后三个贡献用列表: - “First, it proposes two evaluation frameworks for assessing highlight explanations across multiple properties and explanation types, together with a meta-evaluation framework for examining the reliability of existing faithfulness metrics for chain-of-thought explanations.” - “Second, it proposes two techniques that generate natural language explanations designed to be faithful to the model's prediction process while remaining concise, coherent, and helpful for understanding model decisions.” - “Finally, a study of human-AI interaction explores the roles of evidence and explanations in helping users validate model verdicts in a fact-checking setting.” 标题方面,用户说“长文用

    /

    划分小节”。由于这是相对紧凑的文本,使用两个 h2 小节可能比较合适。或者我可以将“目标与贡献”作为一个小节,并使用列表。让我设计一下: ```html

    背景与评估框架

    Language models are increasingly used to support human judgement in complex reasoning tasks, yet their predictions often remain difficult to interpret and validate. Explainability techniques provide insight into model decision-making, helping expose errors, failure modes, and potential bias so users can decide whether to rely on model outputs. Explanations are simplified approximations of a language model's internal workings. The goal is to obtain explanations that reflect the underlying computation as closely as possible (faithfulness), are understandable to users (plausibility), and help them achieve their intended goals (utility). Evaluation protocols are essential for assessing these dimensions.

    研究范围与核心挑战

    This thesis studies highlight explanations, natural language explanations, and chain-of-thought reasoning. These forms offer complementary perspectives on model behaviour and exhibit distinct strengths and limitations across tasks. Producing explanations that satisfy multiple quality dimensions remains challenging, as does evaluating them reliably across forms and settings.

    研究目标与贡献

    This thesis aims to improve people's understanding of models' decision-making processes in complex reasoning tasks.

    • First, it proposes two evaluation frameworks for assessing highlight explanations across multiple properties and explanation types, together with a meta-evaluation framework for examining the reliability of existing faithfulness metrics for chain-of-thought explanations.
    • Second, it proposes two techniques that generate natural language explanations designed to be faithful to the model's prediction process while remaining concise, coherent, and helpful for understanding model decisions.
    • Finally, a study of human-AI interaction explores the roles of evidence and explanations in helping users validate model verdicts in a fact-checking setting.
    ``` 等等,用户说“保留原文所有信息,不得增删改内容”。我给“faithfulness”、“plausibility”、“utility”添加了 `` 标签,并给“First”、“Second”、“Finally”添加了 ``,这只是格式上的强调,并没有改变文本内容,这是被允许的,因为用户明确提到“重要句子可用”。 但我需要重新考虑:加 `` 到这些词上是否算“修改内容”?其实没有,文本内容完全没变,只是加了格式标签。我觉得没问题。 再检查一下是否保留了所有文本内容。让我逐字核对原文: 原文:“Language models are increasingly used to support human judgement in complex reasoning tasks, yet their predictions often remain difficult to interpret and validate. Explainability techniques provide insight into model decision-making, helping expose errors, failure modes, and potential bias so users can decide whether to rely on model outputs. Explanations are simplified approximations of a language model's internal workings. The goal is to obtain explanations that reflect the underlying computation as closely as possible (faithfulness), are understandable to users (plausibility), and help them achieve their intended goals (utility).