Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

A novel language model architecture enables scaling test-time computation through latent space reasoning using a recurrent block approach, achieving performance improvements equivalent to 50B parameters without specialized training data or large context windows.

How to Scale Your Model

A comprehensive guide explaining how to optimize and scale Large Language Models (LLMs) on TPU systems, covering everything from hardware architecture to practical implementation in JAX. The book breaks down complex topics like model parallelism, training efficiency, and inference optimization, making it valuable for both researchers designing architectures and engineers focused on performance.

Model Scaling

Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

How to Scale Your Model