Skip to main content
Technical implementation · AI Search Infrastructure

​
Definition

Post-training refers to the processes applied to a foundation model after initial pre-training — including fine-tuning on task-specific data, reinforcement learning from human feedback (RLHF), and instruction tuning. Post-training shapes how a model responds to queries, follows instructions, and prioritizes different types of information. Post-training is why two models trained on similar data can behave very differently when asked about a brand. Fine-tuning and RLHF introduce biases, priorities, and behavioral patterns that affect citation behavior, tone, and the weighting of different source types. A brand optimized for one model’s base knowledge may not perform equally well across models with different post-training configurations.

Foundation model

Pre-training

Fine-tuning

Inference

Training corpus

​
Relevant Plate Lunch Collective Services

AI SEO AI Search Visibility Assessment