Large Language Models
155 views
Chain-of-Thought Prompting
Quick Definition
Technique encouraging step-by-step reasoning in LLMs
Full Definition
A technique encouraging LLMs to break down complex problems into sequential reasoning steps.
Examples
math problem solving, complex reasoning, debugging
Related Terms
prompt-engineering
few-shot-learning
More Large Language Models Terms
Grounding LLM
Connecting LLM outputs to verifiable external sources
Masked Language Model
Training objective predicting randomly masked input tokens
Positional Encoding
Incorporating token position information into transformers
WordPiece
Subword tokenization splitting words into frequent units
Mixture of Experts
Architecture routing tokens to specialized expert sub-networks
Prompt Caching
Storing key-value states for repeated prompt prefixes