Generative AI and LLM glossary

Data Representation

A structured representation of knowledge that connects entities and their relationships, typically used in AI for information retrieval.

AI Foundations

The method of encoding information about the world into a format that a computer system can utilize.

Knowledge Representation

A repository of structured information and facts for inferencing and decision-making.

Inference

The KV cache stores attention keys and values from already processed tokens so the model does not recompute the whole prefix each step.

Models

A large language model, or LLM, is a neural text model trained on large corpora to predict, generate, transform, and reason over language and code.

Statistical Models

A statistical model that assumes the existence of unobserved variables that influence observed data.

Model Training

The learning rate is a hyperparameter that controls how much to change the model weights in response to the estimated error each time the model weights are updated.

Evaluation

LLM evaluation measures whether a model or AI workflow is accurate, useful, safe, reliable, and cost-effective for a target task.

Inference

A logit is an unnormalized score for a vocabulary item before softmax turns it into a probability.

Model Capabilities

Long context is a product capability: a model or endpoint that accepts a much larger token window than a standard chat SKU.

Training

LoRA, low-rank adaptation, fine-tunes a small set of adapter weights instead of updating every parameter of a base model.

Evaluation Metrics

A method of evaluating how well a specific algorithm models the given data.

Machine Learning

A subset of AI that enables systems to learn from data and improve over time.

Agents

An MCP server exposes tools, data sources, prompts, or workflows to AI clients through the Model Context Protocol.

Optimization

A class of optimization algorithms that use iterative processes to find solutions.

Machine Learning

Learning to learn, where models improve their learning strategies over time.

Model Architecture

A mixture-of-experts model routes each token through a subset of specialized networks instead of activating every parameter.

Machine Learning

Different forms or types of data used in machine learning, such as text, images, or audio.

Agents

Model Context Protocol, or MCP, is a standard interface for connecting AI applications to external tools, data sources, and context providers.

Machine Learning Operations

The process of making a trained machine learning model available for use in a production environment.

Model Assessment

The process of assessing a trained model's performance using various metrics.

Machine Learning

The ability of a machine learning model to perform well on unseen data.

Ethics

The degree to which a human can understand the cause of a decision made by a model.

Machine Learning

A technique used to prevent overfitting by adding a penalty for larger coefficients in a model.