Learn Before
Strategic Model Improvement
A research lab has a large language model and a fixed budget for the next improvement cycle. They are weighing two options:
- Use the budget to acquire and process a new dataset that is ten times larger than their current one.
- Use the budget to fund a team of engineers for several months to experiment with novel, unproven architectural changes to the model, using the existing dataset.
Based solely on the established principle that describes how a model's final test performance relates to the amount of training data, which of these two strategies represents a more predictable path to achieving a lower test loss? Justify your reasoning.
0
1
Tags
Ch.2 Generative Models - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A machine learning team is training a series of language models. They systematically increase the size of the training dataset for each new model and record the final test loss. When they plot the test loss versus the dataset size on a graph where both axes use a logarithmic scale, they observe the points form a nearly straight, downward-sloping line. What is the most valid interpretation of this trend?
Three Phases of LLM Scaling with Dataset Size
Strategic Model Improvement
Interpreting Training Anomalies
Empirical Power Law for LLM Loss vs. Dataset Size (D)