ACL Findings2025
How do LLMs' Preferences Affect Event Argument Extraction? CAT: Addressing Preference Traps in Unsupervised EAE
Yunhao Wei, Kai Shuang, Zhiyi Li, Chenrui Mao
Abstract
Large Language Models (LLMs) have significantly improved the performance of unsupervised Event Argument Extraction (EAE) tasks. However, the prevalent preferences in LLMs severely hinder their effectiveness in EAE, leading to what we term preference traps, namely, the prior knowledge trap, the sycophancy hallucination trap, and the output contradiction trap. Existing methods often fall into these traps due to low-quality prior knowledge, ambiguous instructions, and contradictory outputs. To address this issue, we propose Choose-After-Think (CAT), an unsupervised EAE framework based on LLMs. CAT divides the task into two stages: identification of event information (think stage) and selection of arguments from a candidate argument set for template filling (choose stage). Both stages employ countermeasures to address these preference traps, while the choose stage's completely constrained outputs satisfy EAE's structured-output requirements. Experimental results demonstrate that CAT (based on the local 7B model, zero-shot setting) matches the performance of the best DeepSeek-R1 APIaccessible model, with a significantly lower time cost. 1
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
Builds on14
- LLM Evaluators Recognize and Favor Their Own GenerationsArjun Panickssery, Samuel R. Bowman, Shi FengNeurIPS 2024 · 865 citations
- Towards Understanding Sycophancy in Language ModelsMrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud et al.ICLR 2024 · 762 citations
- A Joint Neural Model for Information Extraction with Global FeaturesYing Lin, Heng Ji, Fei Huang, Lingfei WuACL 2020 · 376 citations
- Code4Struct: Code Generation for Few-Shot Event Structure PredictionXingyao Wang, Sha Li, Heng JiACL 2023 · 40 citations
- Do Androids Laugh at Electric Sheep? Humor "Understanding" Benchmarks from The New Yorker Caption ContestJack Hessel, Ana Marasovic, Jena D. Hwang, Lillian Lee et al.ACL 2023 · 29 citations
Related papers
- Extracting Events Like Code: A Multi-Agent Programming Framework for Zero-Shot Event ExtractionQuanjiang Guo, Sijie Wang, Jinchuan Zhang, Ben Zhang et al.AAAI 2026
- Thinking about how to extract: Energizing LLMs' emergence capabilities for document-level event argument extractionKai Shuang, Ji Zhou, Qiwei Wang, Jinyu GuoACL Findings 2024 · 6 citations
- Rule-Guided Extraction: A Hierarchical Rule Optimization Framework for Document-Level Event Argument ExtractionYue Zuo, Yuxiao Fei, Wanting Ning, Jiayi Huang et al.EMNLP Findings 2025 · 2 citations
- Debate as Optimization: Adaptive Conformal Prediction and Diverse Retrieval for Event ExtractionSijia Wang, Lifu HuangEMNLP Findings 2024 · 5 citations
- REAR: Reinforced Reasoning Optimization for Event Argument Extraction with Relation-Aware SupportJianwen Luo, Yu Hong, Shuai Yang, Jianmin YaoEMNLP Findings 2025
