System Report for CCL25-Eval Task 5: New Dataset and LoRA-Fine-Tuned Qwen2.5

📄 arXiv: 2606.12392v1 📥 PDF

作者: Haotao Xie

分类: cs.CL, cs.AI

发布日期: 2026-06-10


💡 一句话要点

提出PoetryQwen以解决古典诗歌翻译与情感理解问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 古典诗歌 语言模型 情感理解 数据集构建 低秩适应

📋 核心要点

  1. 现有研究将古典诗歌的欣赏任务视为一般领域问题,忽略了其独特性,导致精准翻译和情感理解的研究不足。
  2. 本文通过将任务分解为术语解释、语义解释和情感推断,构建了专门针对古典诗歌的高质量数据集,并提出了PoetryQwen模型。
  3. 实验结果显示,PoetryQwen在基准测试中得分为0.757,较基线模型提升了9.7%,有效提升了古典诗歌的翻译和情感理解能力。

📝 摘要(中文)

近年来,大型语言模型在古典汉语翻译和古典诗歌生成领域取得了显著进展。然而,针对古典诗歌的精准翻译和情感语义理解的领域特定研究仍然有限。现有研究往往将诗歌欣赏任务视为一般领域问题,忽视了其独特特征,且高质量的领域特定数据集极为稀缺。为了解决这一限制,本文将任务分解为术语解释、语义解释和情感推断三个子任务,并基于多个开源数据集进行数据清洗和对齐,构建了包含49,404对高质量指令-响应对的古典汉语诗歌指令对数据集(CCPoetry-49K)。随后,本文通过低秩适应(LoRA)对Qwen2.5-14B模型进行微调,提出了领域专用的语言模型PoetryQwen。实验结果表明,PoetryQwen在CCL25-Eval Task 5基准测试中取得了0.757的得分,相较于基线模型Qwen2.5-14B-Instruct(0.690)提升了9.7%。这些发现表明,PoetryQwen显著提高了古典诗歌的精准翻译和情感理解能力。

🔬 方法详解

问题定义:本文旨在解决古典诗歌翻译和情感理解的不足,现有方法未能充分考虑诗歌的独特特征,导致效果不佳。

核心思路:通过将诗歌欣赏任务分解为术语解释、语义解释和情感推断,构建专门的数据集,并利用LoRA对Qwen2.5进行微调,以优化模型性能。

技术框架:整体流程包括数据清洗与对齐、数据集构建、模型微调和性能评估。主要模块包括CCPoetry-49K数据集和PoetryQwen模型。

关键创新:提出了领域专用的PoetryQwen模型,利用LoRA技术进行微调,显著提升了古典诗歌的翻译和情感理解能力,与现有方法相比,具有更好的适应性和准确性。

关键设计:在数据集构建中,进行了严格的数据清洗和对齐,确保了数据的高质量;在模型微调中,采用了适合古典诗歌特征的损失函数和网络结构设计。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,PoetryQwen在CCL25-Eval Task 5基准测试中得分为0.757,较基线模型Qwen2.5-14B-Instruct的0.690提升了9.7%,显著提高了古典诗歌的翻译和情感理解能力。

🎯 应用场景

该研究的潜在应用领域包括古典文学的翻译、教育以及文化传播等。通过提升古典诗歌的翻译和情感理解能力,能够更好地促进古典文化的传承与发展,具有重要的实际价值和未来影响。

📄 摘要(原文)

Recently, large language models (LLMs) have achieved promising progress in the fields of classical Chinese translation and the generation of classical poetry. However, domain-specific research on precise translation and affective-semantic understanding of classical poetry remains limited. The main challenge is that most studies treat the poetic appreciation task as a general-domain problem, neglecting the distinctive features of poetic appreciation, while high-quality and domain-specific datasets are extremely limited. To address this limitation, we decompose the task into three subtasks: term interpretation, semantic interpretation, and emotional inference. Based on multiple open-source datasets, we perform data cleansing and alignment to construct the Classical Chinese Poetry Instruction Pair Dataset (CCPoetry-49K), which comprises 49,404 high-quality instruction-response pairs explicitly optimized for this domain. We then propose a domain-specialized LLM, called PoetryQwen, by applying Low-Rank Adaptation (LoRA) to fine-tune the Qwen2.5-14B model. Experimental results on the CCL25-Eval Task 5 benchmark demonstrate that PoetryQwen achieves a score of 0.757, representing a 9.7% improvement over the Qwen2.5-14B-Instruct baseline (0.690). These findings clearly indicate that PoetryQwen significantly enhances performance in precise translation and emotional understanding of classical poetry. We present new dataset and methodological considerations intended to support the domain-specific optimization of LLMs.