(5) Techniques to improve generative AI model output

A prompting technique of asking the model to complete a task with no prior examples.

Zero-Shot

A prompting technique of providing the model with one example to learn from.

One-Shot

A prompting technique of giving the model multiple examples to learn from.

Few-shot

A prompting technique of assigning a persona to the model to influence its style, tone, and focus

Role

A prompting technique of engaging in a back and forth conversation with the AI

Prompt chaining

Connecting the AI's output to verifiable sources of information.

Grounding

What does RAG standfor?

Retrieval-Augmented Generation

The LLM retrieves relevant information from external sources using tooling.

Retrieval

The retrieved information is incorporated into the prompt to the LLM.

Augmentation

The LLM processes the prompt and generates a response.

Generation

The LLM can repeat on the retrieval process as necessary.

Iteration

A prompt engineering technique that allows the LLM to reason and take action on a user query

ReAct (reason and act)

A prompt engineering technique that guides a LLM through a problem-solving process by providing examples with intermediate reasoning steps.

CoT (chain-of-thought)

A prompt engineering technique that uses prompting to guide the AI model to generate, modify, or interpret other prompts.

Metaprompting

Saving prompts as templates for repeated use.

Reusing prompts

Continuing conversations within the same chatbot to maintain context.

Leveraging prompt chaining

Storing specific information for the model to use consistently.

Using saved info

Personalized AI assistants within Gemini. They provide personalized responses tailored to specific instructions. They also streamline workflows like templates, prompts, and guided interactions.

Gems

Settings that influence the AI model's behavior, allowing for customized results.

Sampling Parameters

This parameter represents meaningful chunks of text (like words and punctuation).

Token Count

This parameter controls the "creativity" or randomness of the model's word choices during text generation.

Temperature

This parameter is the cumulative probability of the most likely tokens considered during text generation. This is another way to control the randomness of the model's output.

Top-p

These parameter allows you to filter out potentially harmful or inappropriate content from the model's output.

Safety Settings

This parameter determines the maximum length of the generated text.

Output length