From 9c660077b427f67066a2303eb84fd2077f5baf04 Mon Sep 17 00:00:00 2001 From: medusa Date: Sat, 18 Nov 2023 04:16:38 +0000 Subject: [PATCH] Update docs/llm/Effective-LLM-Prompting.md --- docs/llm/Effective-LLM-Prompting.md | 42 +++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/docs/llm/Effective-LLM-Prompting.md b/docs/llm/Effective-LLM-Prompting.md index a3c19c2..96954a9 100644 --- a/docs/llm/Effective-LLM-Prompting.md +++ b/docs/llm/Effective-LLM-Prompting.md @@ -62,6 +62,37 @@ This guide is crafted to empower developers and enthusiasts in creating effectiv - ❌ "Explain machine learning." - ✅ "Write a comprehensive explanation of machine learning for a layman, including practical examples, without using jargon." +## 💡 Practical Application: Iterating on Prompts Based on LLM Responses + +This section offers practical strategies for refining prompts based on the responses from Language Learning Models (LLMs), which is crucial for achieving the most accurate and relevant outputs. + +### 🔄 Iterative Refinement Process +- **Initial Evaluation**: Critically assess if the LLM's response aligns with the prompt's intent. +- **Identify Discrepancies**: Locate areas where the response differs from the expected outcome. +- **Adjust for Clarity**: Refine the prompt to clarify the expected response. +- **Feedback Loop**: Use the LLM's output to iteratively adjust the prompt for better accuracy. + +### 📋 Common Issues & Solutions +- **Overly Broad Responses**: Specify the scope and depth required in the prompt. +- **Under-Developed Answers**: Ask for explanations or examples to enrich the response. +- **Misalignment with Intent**: Clearly state the purpose of the information being requested. +- **Incorrect Assumptions**: Add information to the prompt to correct the LLM's assumptions. + +### 🛠 Tools for Refinement +- **Contrastive Examples**: Use 'do's and 'don'ts' to clarify task boundaries. +- **Sample Outputs**: Provide examples of desired outputs. +- **Contextual Hints**: Embed hints in the prompt to guide the LLM. + +### 🎯 Precision in Prompting +- **Granular Instructions**: Break down tasks into smaller steps. +- **Explicit Constraints**: Define clear boundaries and limits for the task. + +### 🔧 Adjusting Prompt Parameters +- **Parameter Tuning**: Experiment with verbosity, style, or tone settings. +- **Prompt Conditioning**: Prime the LLM with a series of related prompts before the main question. + +Implementing these strategies can significantly improve the effectiveness of your prompts, leading to more accurate and relevant LLM outputs. + ## 🔚 Conclusion This guide is designed to help refine your prompt crafting skills, enabling more effective and efficient use of LLMs for a range of applications. @@ -116,3 +147,14 @@ This taxonomy provides a hierarchical framework for categorizing educational obj - **Creating**: Innovate and formulate new concepts or products. - Initiate and develop original creations or ideas that enhance or extend existing paradigms. +# Latent Content + +This term refers to the reservoir of knowledge, facts, concepts, and information that is integrated within a model and requires activation through effective prompting. + +- **Training Data**: Source of latent content derived exclusively from the data used during the model's training process. +- **World Knowledge**: Broad facts and insights pertaining to global understanding. +- **Scientific Information**: Detailed data encompassing scientific principles and theories. +- **Cultural Knowledge**: Insights relating to various cultures and societal norms. +- **Historical Knowledge**: Information on historical events and notable individuals. +- **Languages**: The structural elements of language, including grammar, vocabulary, and syntax. +