In a June 12, 2024 paper researchers from Tencent AI and the Harbin Institute of Technology introduced TasTe, a method for teaching large language models (LLMs) to translate through self-reflection.

2024年6月12日，腾讯人工智能和哈尔滨工业大学的研究人员在发表的论文中提到了一种教大语言模型（LLMs）通过自我反思进行翻译的方法，即TasTe。

The key idea is to enable LLMs to generate preliminary translations (i.e., drafts), self-evaluate their own translations, and make refinements based on the evaluation.

TasTe的主要做法是让LLMs先初步翻译（即生成草稿） ，然后对译文自我评估，并根据评估完善译文。

The researchers explained that LLMs have shown exceptional performance across various natural language processing tasks, including machine translation (MT). However, their translations still do not match the quality of supervised neural machine translation (NMT) systems.

研究人员解释说，大语言模型在各种自然语言处理任务中表现出色，包括机器翻译（MT）。然而，他们的翻译仍然无法与有监督模式的神经机器翻译（NMT）系统的质量相匹配。

To address this, the authors proposed the TasTe framework (translating through self-reflection), which improves the translation capabilities of LLMs by incorporating a self-reflection process.

鉴此，作者提出TasTe，意在通过加入自我反思提高LLMs的翻译能力。

This process consists of two stages. In the first stage, LLMs are prompted to generate preliminary translations (i.e. drafts) while simultaneously making quality predictions for these translations. The quality predictions can be in the form of labels like “good,” “medium,” and “bad” or scores ranging from 0 to 100. This self-assessment step allows the models to evaluate the quality of their own outputs.

自我反思包含两个阶段。第一阶段，LLMs经提示生成初步翻译，同时对这些翻译进行质量预测。质量预测可以用“好”、“中”和“坏”的标签进行标注，或者给出0到100的评分。这种自我评估的步骤可以让模型对其译文质量进行评价。

In the second stage, LLMs refine these preliminary translations based on the quality predictions in the first stage to produce final translations. According to Xuebo Liu, Assistant Professor at Harbin Institute of Technology, speaking to Slator, low-quality drafts with severe errors undergo extensive modifications, medium-quality drafts with minor errors receive moderate adjustments, and high-quality drafts with minimal or no errors require little to no changes. “By equipping models to tailor their modifications to the draft quality, we effectively rectify conspicuous errors and prevent the misguidance of error propagation that could otherwise compromise originally accurate translations, thereby safeguarding the overall translation quality,” he added.

第二阶段， LLMs在初翻的基础上精进译文，产出最终译文。哈尔滨工业大学助理教授刘雪波在接受Slator采访时表示：质量预测差，错误多的初稿会被大幅修改；质量预测中等，错误较轻的只需适当修改；预测质量高，错误很少甚至没有的，几乎不需要修改。他补充说：“通过让模型根据草稿质量调整修改，我们可以有效地纠正明显的错误，并防止错误扩散，进而保护整体的翻译质量。这种扩散如果不加以控制，可能会影响原本准确的翻译。

This entire process can be seen as a form of self-reflection, mirroring the common “try-evaluate-improve” approach humans use when handling complex tasks to execute them more effectively, Liu said.

他还说，整个过程可以看作是一种自我反思，就像人类为更有效处理复杂任务时经常会采取“尝试-评估-改进”的方式。

原文网址：

How to Teach Large Language Models to Translate Through Self-Reflection – Slator