用两阶段微调提升大模型多语言指代消解性能，夺冠测试集
Closing the Gap at CRAC 2026: Two-Stage Adaptation for LLM-Based Multilingual Coreference Resolution
- 分两阶段微调Gemma-3-27b，先用多语言基适配器再加数据集适配器
- 在官方测试集上平均CoNLL F1达74.32，位居大模型赛道第一
- 适合关注多语言指代消解与模型适配策略的研究者
我们提交了针对2026年计算参考、回指与指代（CRAC 2026）共享任务大模型赛道的方案。在官方测试集上，系统平均CoNLL F1得分为74.32，位列大模型赛道第一，总体排名第三。系统基于Gemma-3-27b模型，采用两阶段微调策略：先使用多语言基础适配器，再接入数据集特定适配器。通过头词表示提及跨度，并采用类XML格式结合局部重索引进行标注，支持迭代式文档处理。该设计在多种语言、文档长度及标注规范下均表现稳健。
原文摘要 · Abstract (English)
We present our submission to the LLM track of the 2026 Computational Models of Reference, Anaphora and Coreference (CRAC 2026) shared task. With an average CoNLL F1 score of 74.32 on the official test set, our system ranked first in the LLM track, and third overall. Our system is based on the Gemma-3-27b model, fine-tuned using a two-stage strategy with a multilingual base adapter followed by dataset-specific adapters. We represent mention spans by their headword using an XML-inspired format with local reindexing and annotate documents iteratively. These design choices proved effective across languages, document lengths, and annotation guidelines.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。