RLHF
Quick Definition
Aligning LLMs with human preferences using reinforcement learning
Full Definition
Reinforcement Learning from Human Feedback aligning language models with human preferences.
Examples
ChatGPT alignment, preference learning, safety tuning
Related Terms
reinforcement-learning
reward-model
alignment
More Large Language Models Terms
Grounded Generation
Generating outputs faithful to provided source documents
LoRA
Parameter-efficient fine-tuning using low-rank matrix decomposition
Guardrails LLM
Frameworks for monitoring and controlling LLM I/O
Flash Attention
Memory-efficient attention using GPU SRAM block computation
Sliding Window Attention
Attention limiting tokens to nearby window only
WordPiece
Subword tokenization splitting words into frequent units