ContrastWSD: Enhancing Metaphor Detection with Word Sense Disambiguation
RoBERTa-based model ContrastWSD enhances metaphor detection by integrating Word Sense Disambiguation, outperforming other methods.
SensoryT5: Integrating Sensory Knowledge for Enhanced Emotion Classification with T5
SensoryT5 integrates sensory information into the T5 model to enhance fine-grained emotion classification, showcasing improved performance and highlighting the potential of neuro-cognitive data in NLP.
PhoGPT: Vietnamese Generative Pre-training Model
Vietnamese PhoGPT models offer state-of-the-art performance and open-source availability for language generation tasks.
LoRA-SP: Streamlined Partial Parameter Adaptation for Efficient Fine-Tuning of Large Language Models
LoRA-SPは、大規模言語モデルの効率的なファインチューニングを実現する革新的な手法であり、計算リソースとメモリ要件を大幅に削減しながら高い性能を維持します。
Two-stage LLM Fine-tuning with ProMoT Framework for Enhanced Generalization
Fine-tuning with ProMoT reduces format specialization and enhances generalization in LLMs.
Efficient Question Answering in Spanish with Language Model Knowledge Distillation
Developing compressed language models through knowledge distillation for efficient question answering in Spanish.
Contrastive Sentence Representation Learning: Understanding and Optimization
Effective contrastive losses in Sentence Representation Learning depend on three key components: Gradient Dissipation, Weight, and Ratio.