Train-to-Test (T2) Scaling Laws
T(
Active
A framework for jointly optimizing LLM training and inference costs.
Train-to-Test (T2) Scaling Laws
A framework for jointly optimizing LLM training and inference costs.
About
Train-to-Test (T2) scaling laws is a research framework developed by academics at the University of Wisconsin-Madison and Stanford University. It addresses the disconnect between traditional LLM pretraining scaling laws (optimizing for training cost) and real-world inference needs (where techniques like repeated sampling increase accuracy but also cost). T2 jointly optimizes a model's parameter size, training data volume, and the number of test-time inference samples to achieve compute-optimal performance for reasoning-heavy applications.
Tags
Performance
Company Timeline
No timeline data for this period
Discussion (0)
Join the discussion
No comments yet. Be the first to share your thoughts!
Frequently Asked Questions
What does Train-to-Test (T2) Scaling Laws do?
Train-to-Test (T2) scaling laws is a research framework developed by academics at the University of Wisconsin-Madison and Stanford University. It addresses the disconnect between traditional LLM pretraining scaling laws (optimizing for training cost) and real-world inference needs (where techniques like repeated sampling increase accuracy but also cost). T2 jointly optimizes a model's parameter size, training data volume, and the number of test-time inference samples to achieve compute-optimal p…
Where is Train-to-Test (T2) Scaling Laws headquartered?
Train-to-Test (T2) Scaling Laws is headquartered in Stanford, California, USA.
What industry does Train-to-Test (T2) Scaling Laws operate in?
Train-to-Test (T2) Scaling Laws operates in Artificial Intelligence, Machine Learning, Computer Science.
Contact Info
Location
Stanford, California, USA
Similar Startups