何时需要为ChatGPT使用思维链提示？
When do you need Chain-of-Thought Prompting for ChatGPT?
AI总结：
本文研究思维链提示在ChatGPT上的有效性，发现其对算术推理等任务不再有效，并揭示指令微调中的过拟合偏差及预训练数据集泄漏风险。
摘要：
思维链（CoT）提示能够有效激发大型语言模型（LLMs）进行复杂的多步推理。例如，仅需在MultiArith数据集的每个输入查询中添加CoT指令“让我们逐步思考”，GPT-3的准确率就能从17.7%提升至78.7%。然而，目前尚不清楚CoT在更新的指令微调（IFT）LLMs（如ChatGPT）上是否仍然有效。令人惊讶的是，在ChatGPT上，CoT对某些任务（如算术推理）不再有效，但在其他推理任务上仍然有效。此外，在前一类任务上，ChatGPT通常能达到最佳性能，并且即使未被指示，也能生成CoT。因此，ChatGPT很可能已经在这些任务上使用CoT进行了训练，从而记住了该指令，因此在应用于相同查询时，即使没有CoT，也会隐式地遵循该指令。我们的分析反映了IFT中引入的指令可能存在的过拟合/偏差风险，这在训练LLMs中变得越来越普遍。此外，它还表明预训练配方可能存在泄漏，例如，可以验证某个数据集和指令是否被用于训练ChatGPT。我们的实验报告了ChatGPT在多种推理任务上的新基线结果，并为LLM的画像分析、指令记忆和预训练数据集泄漏提供了新颖的见解。
英文摘要：
Chain-of-Thought (CoT) prompting can effectively elicit complex multi-step reasoning from Large Language Models~(LLMs). For example, by simply adding CoT instruction ``Let's think step-by-step'' to each input query of MultiArith dataset, GPT-3's accuracy can be improved from 17.7\% to 78.7\%. However, it is not clear whether CoT is still effective on more recent instruction finetuned (IFT) LLMs such as ChatGPT. Surprisingly, on ChatGPT, CoT is no longer effective for certain tasks such as arithmetic reasoning while still keeping effective on other reasoning tasks. Moreover, on the former tasks, ChatGPT usually achieves the best performance and can generate CoT even without being instructed to do so. Hence, it is plausible that ChatGPT has already been trained on these tasks with CoT and thus memorized the instruction so it implicitly follows such an instruction when applied to the same queries, even without CoT. Our analysis reflects a potential risk of overfitting/bias toward instructions introduced in IFT, which becomes more common in training LLMs. In addition, it indicates possible leakage of the pretraining recipe, e.g., one can verify whether a dataset and instruction were used in training ChatGPT. Our experiments report new baseline results of ChatGPT on a variety of reasoning tasks and shed novel insights into LLM's profiling, instruction memorization, and pretraining dataset leakage.
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。
