Pretraining
The broad first phase in which a model learns patterns from a very large dataset.
Full definitionGlossary
The vocabulary behind model improvement.
Short definitions establish orientation. Full entries explain why each concept matters and where it appears in practice.
Try “reward,” “environment,” or “training.”
13 terms
Post-training
Training after pretraining that shapes a model into a useful, reliable assistant or agent.
Full definitionSupervised fine-tuning
Training on curated examples that demonstrate desired inputs and outputs.
Full definitionData annotation
The process of adding labels, judgments, corrections, or other structure to data used for training or evaluation.
Full definitionPreference data
Human or model judgments that compare outputs or score their quality.
Full definitionRLHF
Reinforcement learning from human feedback used to align model behavior with human preferences.
Full definitionReward model
A learned scorer that predicts how desirable a model output is.
Full definitionReinforcement learning
Learning behavior through actions, feedback, and rewards in an environment.
Full definitionBenchmark
A standardized set of tasks and scoring rules used to compare model performance.
Full definitionEvaluation
The systematic measurement of a model's capabilities, quality, and failure modes.
Full definitionSynthetic data
Artificially generated examples used for training or evaluation.
Full definitionAgent environment
A controlled world in which an AI agent observes, acts, and receives feedback.
Full definitionInference
The process of running a trained model to produce an output or take an action.
Full definition