Reproduce “Resolving the Imbalance Issue in Hierarchical Disciplinary Topic Inference via LLM-based Data Augmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xunxin Cai, Meng Xiao, Zhiyuan Ning, Yuanchun Zhou
A framework named HiPSTG employs Llama V1 to augment scientific research proposals, mitigating data imbalance in hierarchical disciplinary topic inference. This method enhances classification performance by 8% in MicroF1 and supports fairer AI-assisted expert assignments through generated high-quality data for underrepresented disciplines.
Key insights
- Large Language Models (LLMs) can effectively generate high-quality, specialized scientific text, making them suitable for data augmentation in complex, domain-specific hierarchical classification tasks.
- Integrating hierarchical structural information and specific keywords into prompt design is crucial for guiding LLMs to produce diverse, semantically rich, and relevant synthetic data for minority classes.
- Addressing data imbalance in hierarchical scientific text classification directly contributes to enhancing fairness and reducing bias in AI-assisted scientific administration processes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.