NLP Task
Natural Language Processing (NLP) research currently focuses on enhancing Large Language Models (LLMs) for a wider range of tasks, including improved long-context processing, reliable benchmark creation using synthetic data, and seamless integration of generation and retrieval capabilities. Active research areas involve developing efficient frameworks for handling extensive input sequences within memory constraints, evaluating the effectiveness of LLMs across diverse and challenging benchmarks (including those for specialized domains like finance and law), and mitigating issues like data contamination and hallucination. These advancements are crucial for improving the reliability and applicability of LLMs in various real-world applications, from legal tech to healthcare and beyond.
Papers
Beyond Memorization: The Challenge of Random Memory Access in Language Models
Tongyao Zhu, Qian Liu, Liang Pang, Zhengbao Jiang, Min-Yen Kan, Min Lin
SemEval-2024 Shared Task 6: SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes
Timothee Mickus, Elaine Zosa, Raúl Vázquez, Teemu Vahtola, Jörg Tiedemann, Vincent Segonne, Alessandro Raganato, Marianna Apidianaki
SiLLM: Large Language Models for Simultaneous Machine Translation
Shoutao Guo, Shaolei Zhang, Zhengrui Ma, Min Zhang, Yang Feng
Comparing Specialised Small and General Large Language Models on Text Classification: 100 Labelled Samples to Achieve Break-Even Performance
Branislav Pecher, Ivan Srba, Maria Bielikova