Large Language Models

Lab 3: Conceptual-Overview

Agenda

1. Large Language Models

2. Zero shot

3. Few shot

4. Chain of thought

Large Language Models (LLMs)

Large Language Models (LLMs) are types of artificial intelligence algorithms designed to understand, generate, and sometimes translate human language. They are trained on vast amounts of text data, allowing them to learn a wide variety of linguistic patterns, syntax, and semantics. This extensive training enables them to perform a range of language-based tasks that were previously challenging for machines.

Examples of Large Language Models:

• OpenAI’s GPT (Generative Pre-trained Transformer) series: Models like GPT-3 have demonstrated capabilities in generating human-like text based on the prompts they receive.

• Google’s BERT (Bidirectional Encoder Representations from Transformers): Focused on improving search results by understanding the context of search queries.

• Facebook’s RoBERTa (Robustly optimized BERT approach): An optimized version of BERT that improves language understanding.

Zero-Shot Learning

Zero-shot learning refers to the ability of a model to correctly perform a task without having seen any examples from that specific task during training.

Resource: https://saturncloud.io/blog/breaking-the-data-barrier-how-zero-shot-one-shot-and-few-shot-learning-are-transforming-machine-learning/

Few-Shot Learning

Few-shot learning involves training a model on a very limited amount of data about a specific task. Unlike zero-shot learning, few-shot learning requires some examples (though still very few) of the task to guide the model.

Resource: https://medium.com/ubiai-nlp/step-by-step-guide-to-mastering-few-shot-learning-a673054167a0

Chain of Thought

Chain of thought prompting is a technique used to encourage a model to ‘think aloud’ or step through its reasoning process as it works toward a solution.

Resource: https://www.promptingguide.ai/techniques/cot