🏢 Institute of Computing and Intelligence, Harbin Institute of Technology, Shenzhen, China
SelectIT: Selective Instruction Tuning for LLMs via Uncertainty-Aware Self-Reflection
·3120 words·15 mins·
loading
·
loading
Natural Language Processing
Large Language Models
🏢 Institute of Computing and Intelligence, Harbin Institute of Technology, Shenzhen, China
SelectIT leverages LLMs’ intrinsic uncertainty to efficiently select high-quality instruction tuning data, enhancing model performance without extra resources.