Parameters
Domain: LLM Concepts
🎯 Motivation
Parameters are the learned numbers inside an LLM - weights and biases. More parameters generally mean more capacity to learn complex patterns.
📋 Concrete Example
GPT-3 has 175 billion parameters. GLM-4 has hundreds of billions. Each parameter is a number that was adjusted during training. Together they encode the model’s knowledge.
🔗 Analogy
Knobs on a sound mixing board - each knob (parameter) adjusts something slightly. With 175 billion knobs, you can fine-tune the sound (language) incredibly precisely. Training is finding the right knob settings.