发表机构
University of Naples Federico II; University of Salerno（那不勒斯费德里科二世大学; 萨勒诺大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究文档级模式约束事件论元抽取，结合角色集注入、参数高效监督微调及确定性解码后处理方法，使微调后的中型开放模型在MAVEN-ARG评估中优于GPT基线，最好模型在事件共指级别达42.39% F1值。
AI 中文摘要
事件论元抽取（EAE）通过识别论元跨度并为其分配模式定义的角色，将文档转换为结构化事件记录。文档级EAE具有挑战性，因为触发器和论元之间存在长距离依赖、跨句子上下文以及严格的角色约束，这通常会导致边界错误、角色不确定性以及与受限模式不一致。本文研究了中型开放语言模型是否能在MAVEN-ARG上可靠地执行文档级模式约束EAE。我们的方法包括：在提示中注入角色集以符合模式；使用推理时相同的纯JSON接口进行参数高效的监督微调（LoRA）；以及通过后处理进行确定性解码，该后处理可验证JSON、过滤无效角色、去重论元并将跨度与文档窗口对齐。在官方MAVEN-ARG评估器下，微调后的中型开放模型在提及、实体共指和事件共指评估中优于先前报告的GPT基线；我们最好的模型（Phi-4，14B）在事件共指级别达到了42.39% 的F1值。可在这个https URL上公开获取重现实验的代码。
英文摘要
Event Argument Extraction (EAE) converts documents into structured event records by identifying argument spans and assigning them schema-defined roles. Document-level EAE is challenging due to long-range dependencies between triggers and arguments, cross-sentence context, and strict role constraints, which often lead to boundary errors, uncertainty in roles, and inconsistencies with restricted schemas. In this paper, we study whether mid-sized open LLMs can perform schema-constrained EAE reliably at the document level on MAVEN-ARG. Our approach combines (i) role-set injection in prompts for schema compliance, (ii) parameter-efficient supervised fine-tuning (LoRA) using the same JSON-only interface used at inference, and (iii) deterministic decoding with post-processing that validates JSON, filters invalid roles, de-duplicates arguments, and aligns spans to the document window. Under the official MAVEN-ARG evaluator, fine-tuned mid-sized open models outperform previously reported GPT baselines across mention, entity-coreference, and event-coreference evaluations; our best model (Phi-4, 14B) reaches 42.39\% F1 at the event-coreference level. Code to reproduce experiments is publicly available at https://github.com/dessertlab/EAE/.
CommentsAccepted at ECML PKDD 2026
