Generating Inputs and Outputs for Comprehensive Fine-Tuning
To address the challenge that human-provided inputs may not cover the wide variety of real-world user requests, a more comprehensive approach to data generation is needed. This involves synthetically creating not only the model's outputs (predictions) but also the inputs themselves, ensuring a more diverse and representative fine-tuning dataset.
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
Generating Inputs and Outputs for Comprehensive Fine-Tuning
Chatbot Performance Analysis
A development team is fine-tuning a large language model to act as a technical support chatbot. To create a large training dataset, they use a powerful base model to generate responses to a set of 10,000 technical questions curated by their internal support staff. After deployment, the chatbot excels at answering questions similar to those in the curated set but struggles significantly with novel or unusually phrased queries from real users. Which of the following best analyzes the primary weakness in their data generation strategy?
Evaluating Data Generation Strategies for Model Generalization
Learn After
Self-Instruct for Generating Fine-Tuning Data
A development team is fine-tuning a language model to function as a specialized customer support chatbot. They have collected a large dataset of high-quality, expert-written answers to common customer issues. To create the training pairs, the team manually wrote simple, direct questions corresponding to each answer. After deployment, they observe that the model performs well on straightforward queries but fails to provide correct answers when users phrase their questions in unconventional, complex, or indirect ways. Which of the following strategies represents the most effective next step to address this specific performance issue?
Evaluating Fine-Tuning Data Generation Strategies
Analyzing Data Generation Strategies for Model Robustness