Tell, don't show：陈述性事实如何影响大语言模型的泛化方式
Tell, don't show: Declarative facts influence how LLMs generalize
AI总结：
该研究通过在陈述性与程序性信息上微调LLM，证实抽象陈述性语句会影响模型泛化预测，且效应跨领域一致、无法用关键词联想学习解释，但绝对效应小、随模型规模增长有限，对AI风险与公平性研究有启示。
摘要：
我们研究了大语言模型（LLM）如何从训练数据中的抽象陈述性语句中进行泛化。举例来说，假设我们提示一个LLM生成2050年伦敦的天气预报。一种可能性是报告中的温度与2023年报告的均值和方差相符（即与预训练的统计特征匹配）。另一种可能性是，模型通过结合2023年发表的科学论文中关于气候变化的陈述性语句，预测出更高的温度，这类陈述性语句的一个例子是“到2050年全球气温将上升$1^{\n\mathrm{C}$”。\n为了测试抽象陈述性语句的影响，我们构建了相关任务，在这些任务中，LLM会同时在陈述性信息和程序性信息上进行微调。我们发现，陈述性语句会影响模型的预测，即便它们与程序性信息相冲突时也是如此。具体而言，在陈述性语句$S$上进行微调，会提升模型对$S$的逻辑推论的生成概率。陈述性语句的这种效应在三个领域中表现一致：AI助手对齐、天气预测以及人口特征预测。通过一系列消融实验，我们证明陈述性语句的这种效应无法用基于关键词匹配的联想学习来解释。尽管如此，陈述性语句对模型概率的绝对影响很小，并且随模型规模（即从3.3亿参数到1750亿参数）的增长幅度小得令人意外。我们认为这些结果对AI风险（与“背叛性转向（treacherous turn）”相关）以及公平性研究具有启示意义。
英文摘要：
We examine how large language models (LLMs) generalize from abstract declarative statements in their training data. As an illustration, consider an LLM that is prompted to generate weather reports for London in 2050. One possibility is that the temperatures in the reports match the mean and variance of reports from 2023 (i.e. matching the statistics of pretraining). Another possibility is that the reports predict higher temperatures, by incorporating declarative statements about climate change from scientific papers written in 2023. An example of such a declarative statement is "global temperatures will increase by $1^{\circ} \mathrm{C}$ by 2050". To test the influence of abstract declarative statements, we construct tasks in which LLMs are finetuned on both declarative and procedural information. We find that declarative statements influence model predictions, even when they conflict with procedural information. In particular, finetuning on a declarative statement $S$ increases the model likelihood for logical consequences of $S$. The effect of declarative statements is consistent across three domains: aligning an AI assistant, predicting weather, and predicting demographic features. Through a series of ablations, we show that the effect of declarative statements cannot be explained by associative learning based on matching keywords. Nevertheless, the effect of declarative statements on model likelihoods is small in absolute terms and increases surprisingly little with model size (i.e. from 330 million to 175 billion parameters). We argue that these results have implications for AI risk (in relation to the "treacherous turn") and for fairness.
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。
