Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback.
Hongshen XuZichen ZhuDa MaSituo ZhangShuai FanLu ChenKai YuPublished in: CoRR (2024)
Keyphrases
- domain knowledge
- multiple choice questions
- knowledge base
- knowledge representation
- information retrieval
- reinforcement learning
- expert systems
- knowledge acquisition
- training sessions
- answer questions
- subject matter
- knowledge sharing
- domain experts
- knowledge based systems
- prior knowledge
- learning algorithm
- data mining techniques
- supervised learning
- training samples
- training examples
- artificial neural networks
- background knowledge
- training set
- training phase
- natural language
- multi agent
- neural network