training

Definition

The concrete process of feeding data to a model and updating its weights, typically split into large-scale pre-training followed by fine-tuning or post-training. It dominates the compute budget of most projects, so papers report training cost, data mixture, and schedule as first-class details. Compare inference, which runs a finished model without changing any weights.

Explainers using this term

• Paper Explained: Video Generation Models: A Survey of Post-Training and Alignment — Four Toolboxes for Making Video Models Do What You AskVideo Generation Models: A Survey of Post-Training and Alignment
• Paper Explained: Post-Training Leaves Behavioral Shadows on Unrelated Decisions — capability transfer through one unrelated wordPost-Training Leaves Behavioral Shadows on Unrelated Decisions
• Paper Explained: LIT — Stopping Robots from Cheating on Visual Cues by Giving Vision a Single DoorwayBreaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models

In the textbook

• What Happens When You Make It Bigger — Scaling and Pre-training