CLadder: Assessing Causal Reasoning in Language Models

Paper: "CLadder: Assessing Causal Reasoning in Language Models" (NeurIPS 2023) by Zhijing Jin*, Yuen Chen*, Felix Leeb*, Luigi Gresele*, Ojasv Kamal, Zhiheng Lyu, Kevin Blin, Fernando Gonzalez, Max Kleiman-Weiner, Mrinmaya Sachan, Bernhard Schölkopf. (http://arxiv.org/abs/2312.04350)
<br>
<br>
Abstract: The ability to perform causal reasoning is widely considered a core feature of intelligence. In this work, we investigate whether large language models (LLMs) can coherently reason about causality. Much of the existing work in natural language processing (NLP) focuses on evaluating commonsense causal reasoning in LLMs, thus failing to assess whether a model can perform causal inference in accordance with a set of well-defined formal rules. To address this, we propose a new NLP task, causal inference in natural language, inspired by the "causal inference engine" postulated by Judea Pearl et al. We compose a large dataset, CLadder, with 10K samples: based on a collection of causal graphs and queries (associational, interventional, and counterfactual), we obtain symbolic questions and ground-truth answers, through an oracle causal inference engine. These are then translated into natural language. We evaluate multiple LLMs on our dataset, and we introduce and evaluate a bespoke chain-of-thought prompting strategy, CausalCoT. We show that our task is highly challenging for LLMs, and we conduct an in-depth analysis to gain deeper insight into the causal reasoning abilities of LLMs. Our data is open-sourced at https://github.com/causalNLP/cladder, and our code can be found at https://huggingface.co/datasets/causalnlp/CLadder.

论文：《CLadder：评估大语言模型的因果推理能力》（NeurIPS 2023），作者团队包括金志敬*、陈源*、费利克斯·利布*、路易吉·格雷塞莱*、奥贾斯夫·卡马尔、吕志恒、凯文·布兰、费尔南多·冈萨雷斯、马克斯·克莱曼-韦纳、米尼马亚·萨尚、伯纳德·朔尔科普夫。论文预印本链接：http://arxiv.org/abs/2312.04350

摘要：因果推理能力被普遍视为智能的核心特征之一。本研究旨在探究大语言模型（Large Language Model，LLM）能否对因果关系进行连贯推理。当前自然语言处理（Natural Language Processing，NLP）领域的多数既有研究仅聚焦于评估大语言模型的常识因果推理能力，无法验证模型是否能够遵循一套明确定义的形式化规则开展因果推断。为填补这一研究空白，我们受朱迪亚·珀尔（Judea Pearl）等人提出的因果推断引擎（causal inference engine）启发，提出一项全新的自然语言处理任务——自然语言因果推断。我们构建了包含10000条样本的大型数据集CLadder：基于一系列因果图与三类查询（关联查询、干预查询与反事实查询），通过神谕因果推断引擎生成符号化问题与标准答案，随后将其转化为自然语言形式。我们在该数据集上对多款大语言模型进行了评估，同时提出并验证了一种定制化思维链提示策略CausalCoT。实验结果表明，该任务对大语言模型极具挑战性；我们还开展了深入分析，以进一步揭示大语言模型的因果推理能力。本数据集开源地址为https://github.com/causalNLP/cladder，代码开源地址为https://huggingface.co/datasets/causalnlp/CLadder。