📖 Content moderation glossary
What is an LLM?
In short
A large language model (LLM) is an artificial intelligence model trained on very large amounts of text to understand and generate language; in moderation, it can read content in context and apply rules written in plain language.
LLMs are the technology behind conversational assistants. Their strength for moderation is understanding: they grasp irony, coded insults, context from the article being commented, and they can follow instructions such as 'criticism of our products is allowed, personal attacks on employees are not'.
Example: a commenter writes 'nice job as always' under a post about a delayed train. An LLM can recognise sarcasm and classify it as criticism, which should stay visible, rather than praise or abuse.
LLMs also have limits: they can be inconsistent from one run to another, be fooled by carefully crafted text, and their cost and speed vary. That is why serious moderation systems frame them with rules, thresholds, cross-checks between models and human review. OpenModeration's own moderation AI is cross-checked by other AI models on delicate cases, and new models are added as they come out.