Learning techniques

Large Language Models (LLMs) Concepts

Vidhi Chugh

AI strategist and ethicist

Where are we?

Getting beyond data constraints

 

• Fine-tuning: training a pre-trained model for a specific task

 

• But, what if there is little to no labeled data?

 

• N-shot learning: zero-shot, few-shot, and multi-shot

Transfer learning

• Learn from one task and transfer to related task
• Transferring knowledge from piano to guitar
  • Reading musical notes
  • Understanding rhythm
  • Grasping musical concepts

• N-shot learning
  • Zero-shot - no task-specific data
  • Few-shot - little task-specific data
  • Multi-shot - relatively more training data

Zero-shot learning

• No explicit training
• Uses language understanding and context
• Generalizes without any prior examples

¹ Freepik

Few-shot learning

• Learn a new task with a few examples

• One-shot learning: fine-tuning from one example

• Prior knowledge to answer new question

Multi-shot learning

• Requires more examples than few-shot

 

• Previous tasks, plus new examples

 

• For example, a model trained on Golden Retriever

¹ Freepik

Multi-shot learning

• Model output: Labrador Retriever

 

• Saves time in collecting and labeling data

 

• No compromise on accuracy

¹ Freepik

Building blocks so far

• Data preparation workflow

 

• Fine-tuning

 

• N-shot learning techniques

 

• Next up: pre-training

Let's practice!

Large Language Models (LLMs) Concepts