Researchers investigated methods for improving Chinese metaphor identification across different datasets. They compared four approaches: BERT fine-tuning (BERT-FT), QLoRA-based LLM fine-tuning (LLM-FT), direct zero-shot LLM prompting (LLM-ZS), and zero-shot prompting with a procedural Skill (Skill-ZS). While fine-tuning methods achieved higher accuracy on their native datasets, Skill-ZS demonstrated more stable performance across multiple external datasets, suggesting it as a complementary approach for consistent cross-dataset results. AI
IMPACT This research offers a method to improve the consistency of AI models in identifying metaphors across different data sources.
RANK_REASON The cluster contains an academic paper detailing a new methodology for natural language processing tasks. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
